A RAINFALL-RUNOFF MODELING PROCEDURE FOR IMPROVING ...
Transcript of A RAINFALL-RUNOFF MODELING PROCEDURE FOR IMPROVING ...
'"
WISCONSIN WRD DISTRICT LIBRARY COPY
A RAINFALL-RUNOFF MODELING PROCEDURE FOR IMPROVING ESTIMATES OF T-YEAR (ANNUAU FLOODS FOR SMALL DRAINAGE BASINS
U.S. GEOLOGICAL SURVEY
Water-Resources Investigations 78-7
BIBLIOGRAPHIC DATA 1- Kcpori No. . 2. SHEET
4. Tide and Subtitle
A rainfall-runoff modeling procedure for improving estimates of T-year (annual) floods for small drainage basins
7. Author(s)R. W. Lichty and F. Liscum
9. Performing Organization Name and Address
U.S. Geological Survey Water Resources Division Denver Federal Center, Box 25046 Lakewood, CO 80225
12. Sponsoring Organization Name and Address
U.S. Geological Survey Water Resources Division Denver Federal Center, Box 25046 Lakewood, CO 80225
3. Recipient's Accession No.
5. Report Date
August 19786.
8. Performing Organization RepNo. USGS/WRI 78-7
10. Project/Task/U'ork Unit No
11. Contract/Grant No.
13. Type of Report & Period Covered
Final
14.
15. Supplementary Notes
16. Abstracts Maps depicting the influence of a climatic factor, C, on the magnitude of synthetic T-year (annual) floods were prepared for a large portion of the eastern United States. The climatic factors were developed by regression analysis of flood data generated using a parametric rainfall-runoff model and long-term rainfall records. Map estimates of C values and calibrated values of rainfall-runoff model parameters were used as variables in a synthetic T-year flood relation to compute "map-model" flood estimates for 98 small drainage basins in a six-state study area. Improved esti- ma,tes of T-year floods were computed as a weighted average of the map-model estimate and an observed estimate, with the weights proportional to the relative accuracies of the two estimates. The accuracy of the map-model estimates was appraised by decompos ing components of variance into average time-sampling error associated with the observed estimates and average map-model error. Map-model estimates have an accuracy, in terms of equivalent length of observed record, that ranges from 6 years for the 1.25-year flood up to 30 years for the 50- and 100-year flood._____________________17. Key Words and Document Analysis. 17o. Descriptors
Floods, climatic data, rainfall-runoff relationships, synthetic hydrology, regression analysis, regional analysis
17b. Identifjers/Open-Ended Terms
17c. COSAT1 Field 'Group
18. Availability St^iemcnt
No restriction on distribution
19. Security Class (This Report)
UNCLASS1FIFD20. Security Class (This
Page UNCLASMFIF.D
21. No. of Pages
4422. Price
FORM NTis-55 iKEv. ic-73. LNIXJKM.U HY ANSI A.MJ UNKSCO. THIS FORM MAY BL RtHKODUCED USCOMM-DC EZ6S-I
A RAINFALL-RUNOFF MODELING PROCEDURE
FOR IMPROVING ESTIMATES OF T-YEAR (ANNUAL)
FLOODS FOR SMALL DRAINAGE BASINS
By R. W. Lichty and F. Liscum
U.S. GEOLOGICAL SURVEY
Water-Resources Investigations 78-7
August 1978
UNITED STATES DEPARTMENT OF THE INTERIOR
CECIL D. ANDRUS, Secretary
GEOLOGICAL SURVEY
H. William Menard, Director
For additional information write to:
U.S. Geological SurveyWater Resources DivisionBox 25046, Denver Federal CenterDenver, Colorado 80225
CONTENTS
PageAbstract 1Introduction 2Methods of study 3
Rainfall-runoff model 4Selection of model calibration sites 5Synthesis of annual floods 5Formulation of multiple-regression model 7Regression analysis of synthetic T-year floods rural modelapplications 8
Effects of imperviousness on synthetic floods 11Single-coefficient relation for synthetic T-year floods 12Geographic variability of climatic factor, C 14
Map-model estimates of T-year floods at calibration sites 18Comparison of observed and map-model flood estimates 21
Accuracy and weighting of observed and map-model T-year floods 23Summary and conclusions 31References 33
ILLUSTRATIONS
Figure 1. Schematic outline of model structure, showing components,parameters, variables, and input/output data 4
2. Map showing locations of long-term rainfall, and smallstreams, calibration sites used in study 6
3. Relationship between regression constant, a., andregression coefficient, £> 9 10
Ci ^4. Relationship between the coefficients a., and d. 13
(s Is
5. Contour map showing the geographic variation in the 2-yearrecurrence interval climatic factor, C^ 15
6. Contour map showing the geographic variation in the 25-yearrecurrence interval climatic factor, C~r 16
7. Contour map showing the geographic variation in the100-year recurrence interval climatic factor, ^200 "^
8. Scatter diagrams of observed and map-model T-year floodestimates 22
9. Trends in error-variance components as a function ofrecurrence interval 29
III
TABLES
Page Table 1. Model parameters and variables and their application in
the modeling process 352. Long-term recording rainfall sites used in study 363. Streamflow stations used in study 37-404. Results of multiple-regression analyses for rural model
applications 415. Summary and comparison of multiple-regression (w-2?) and
direct-search (d-s) determinations for the coefficient, a«- 42t'6. Results of the direct-search determinations for the
coefficient d . for effect of impervious area 43
7. Standard errors for the single-coefficient, synthetic floodrelation 44
IV
A RAINFALL-RUNOFF MODELING PROCEDURE FOR IMPROVING ESTIMATES OF
T-YEAR (ANNUAL) FLOODS FOR SMALL DRAINAGE BASINS
By R. W. Lichty and F. Liscum
ABSTRACT
The U.S. Geological Survey rainfall-runoff model is used to synthesize a sample of 550 annual-flood series, that are representative of both rural- and impervious-area model applications, using data from each of 36 long-term recording rainfall sites. A flood-frequency curve is developed for each annual-flood series, and a single-coefficient, regression relation for the 2-, 25-, and 100-year floods is developed for each one of the rainfall sites a generalized definition of the model output as a function of the model param eters for each rainfall site. The site-to-site variability in the magnitude of the coefficient that characterizes the synthetic T-year (annual) flood relation is interpreted as reflecting the spatially varying influence of local climatic factors, C{, 9 on the results of synthesis. Three contour maps that depict the geographic variability of the climatic factor were prepared. Esti mates of the C^ values taken from these maps were used in conjunction with fitted rainfall-runoff model parameters and the synthetic T-year flood rela tion to develop map-model, T-year flood estimates for 98 rural-area streamflow stations located in Missouri, Illinois, Tennessee, Mississippi, Alabama, and Georgia. Comparisons of these flood estimates with those based on observed annual floods show that the map-model estimates are generally lower than the observed estimates for return periods greater than the 2-year recurrence interval. This tendency to underestimate the higher recurrence interval floods was removed by use of an average adjustment factor, B£, and the average accuracy of "unbiased", map-model flood estimates was appraised for the test sample of 98 streamflow stations. The accuracy of the map-model estimates increases rapidly with increasing recurrence interval up to the 10-year inter val, and then reverses its trend and decreases slowly. The map-model flood estimates are more accurate beyond the 10-year recurrence interval than observed estimates based on a harmonic-mean record length of 13.2 years.
Improved T-year flood estimates were computed by weighting observed and m^p-model estimates, and the accuracy of the improved estimates is appraised as a, function of recurrence interval, and in terms of the concept of
equivalent length of record. The map-model estimating procedure yields an equivalent length of record that ranges from a low of about 6 years for the 1.25-year flood, up to an ultimate, maximum level of about 30 years of data for estimating the 50- and 100-year recurrence interval floods.
INTRODUCTION
Historically there has been a deficiency of flood data for drainage basins smaller than about 50 square kilometers. Yet the need for such data is great because accurate and timely estimates of the magnitude and frequency of T-year (annual) floods are an important consideration in the design of drain age structures and the delineation of flood-prone areas. The U.S. Geological Survey, in cooperation with various State, Federal, and local agencies, has undertaken an extensive program of data collection to develop a better knowl edge of the flood-frequency characteristics of small basins.
Many years of annual flood data are required to reliably define a flood- frequency curve. One procedure utilized by the Survey to estimate T-year floods from short records is rainfall-runoff modeling. Rainfall-runoff model ing is undertaken because relatively long records (60 to 70 years) of rainfall are available at many locations throughout the country. Basically, the con cept is that the information contained in the long-term rainfall records can be transformed to streamflow information by synthesizing a long record of annual floods using a calibrated rainfall-runoff model, such as that developed by Dawdy, Lichty, and Bergmann (1972). That is, a short record of observed floods can be effectively lengthened, and the time-sampling error inherent in small samples can be reduced, by the rainfall-runoff modeling process.
A major problem associated with this type of rainfall-runoff model appli cation is the lack of long-term rainfall data at each site for which model- extended data are required or desired. The goal of reducing time-sampling error in observed flood-frequency estimates can, in general, only be achieved if a method of integrating and transferring information from the available long-term rainfall sites is incorporated in the modeling procedure. In addi tion, the accuracy of the transfer mechanism must be appraised to allow a meaningful weighting of observed and modeled flood-frequency estimates in relation to their respective accuracies.
This report describes and evaluates a procedure for computing improved estimates of T-year floods that incorporates a rainfall information transfer mechanism in the form of three maps, and a generalized definition of synthetic T-year flood potential as a function of fitted rainfall-runoff model param eters. The maps depict the geographic variability of a coefficient, £ £, (i = 2-, 25-, and 100-year recurrence intervals) that reflects the influence of local climatic factors on synthetic T-year floods. The climatic factors are derived by analysis of the results of synthesis using the rainfall-runoff model in conjunction with long-term rainfall data and an average daily pattern of potential evapotranspiration.
The generalized definition of synthetic T-year floods facilitates the computation of a "map-model", flood-frequency curve at small-basin calibration sites by using map estimates of the climatic factor, G£, and fitted rainfall- runoff model parameters. Comparisons of observed and map-model, T-year, flood estimates, for a sample of 98 rural-area calibration sites located in Missouri, Illinois, Tennessee, Mississippi, Alabama, and Georgia, show that the map- model estimates are generally too low in relation to observed estimates for recurrence intervals greater than the 2-year or median flood. Average bias correction factors, B£, are determined for the 2-, 25-, and 100-year recur rence intervals, and the accuracy of "unbiased", map-model, T-year flood esti mates is evaluated by decomposing error variance components into average error variance of the map-model estimates, and average time-sampling error variance associated with the observed flood estimates.
The procedure for computing an improved T-year flood estimate involves the computation of a weighted average of the observed and map-model estimates. The weighting factors are developed for each calibration site as a function of the time-sampling variance of the observed estimate and the average variance of the map-model estimate.
METHODS OF STUDY
Many flood-frequency investigations have successfully utilized multiple- regression analysis to explain the variability in the occurrence of floods, and to provide a generalized definition of the magnitude of T-year floods as a function of drainage-basin characteristics. For example, Benson (1962, 1964) and Thomas and Benson (1970) demonstrated that observed T-year flood estimates may be related to various topographic and climatic factors (basin character istics) in the form
bn
where
J = predicted value of T-year flood,
X1 to X = basin and climatic characteristics (drainage area, slope, precipitation intensity, and so forth),
a =* regression constant, and
&- to b =* regression coefficients.In
In an analogous manner, multiple- regression analysis should be useful in explaining the variability of synthetic floods derived by rainfall-runoff modeling using a particular long-term rainfall record. It should also provide
a generalized definition of the magnitude of synthetic T-year floods, as a function of rainfall-runoff model parameters, for each individual rainfall site. Study and interpretation of the regression results should provide insight into how the influence of climate manifests itself in the magnitude, interaction and geographic variability of the regression constant and coeffi cients, a, b 19 b 09 ...b .
j. ci YL
Rainfall-Runoff Model
The rainfall-runoff model used in this study is a simplified, conceptual, bulk-parameter, mathematical model of the surface-runoff component of flood- hydrograph response to storm rainfall. (Dawdy, and others, 1972.) The con tribution of base flow, interflow, and quick return flow to the flood hydro- graph are not considered because they are generally negligible. The model deals with three components of the hydrologic cycle antecedent soil moisture, storm infiltration, and surface-runoff routing. The first component simulates soil-moisture conditions at the onset of a storm period through the application of moisture-accounting techniques on a daily cycle. Estimates of daily rain fall, evaporation, and initial values of the moisture storage variables are elements used in this component. The second component involves an infiltration equation (Philip, 1954) and certain assumptions by which rainfall excess is determined on a 5-minute accounting cycle from storm-period rainfall. Storm rainfall may be defined at 5-, 10-, 15-, 30-, or 60-minute intervals, but loss rates and rainfall excess amounts are computed at 5-minute intervals. The third component transforms the simulated time pattern of rainfall excess into a flood hydrograph by translation and linear storage attenuation (Clark, 1945.) The structure of the model is shown in figure 1 (following). Table 1 (page 35) summarizes the model parameters and their application in the modeling process; (all tables are at end of report). For a more complete description of the model see Dawdy, Lichty, and Bergmann (1972).
ANTECE ACCOl
INPUT
daily rainfall, evaporation.
initial values of BMS and
SMS
DENT SOIL- MOISTURE JNTING COMPONENT
Parameters
BMSM, RR, EVC, DRN,
Variables
BMS, SMS,
INPUT
storm rainfall
OUTPUT- INPUT
BMS , SMS
INFILTRATION SURFACE-RUNOFF COMPONENT ROUTING COMPONENT
Parameters
KSAT. PSP, RGf, BMSM
Variables
BMS, SMS, FR
OUTPUT -INPUTrainfall excess
Parameters
KSW, TC, TP/TC
Variables
SW,
OUTPUT
flood hydrograph
FIGURE 1r ~ Schematic outline of the rainfal I runoff model structure, showing components, parameters, variables, and input output data, flow is left to right.
The model and an automated procedure for determining optimal parameter values are included in both FORTRAN and PL/1 computer language programs, described by Carrigan (1973). The programs provide for the input and storage of observed data, the simulation of flood hydrograph response to storm rain fall, and multistep optimization of model parameters to minimize the error between observed and computed flood peaks.
Selection of Model Calibration Sites
Model calibration results were assembled for sites in Missouri (Hauth, 1974), Alabama (McCain, 1974), Mississippi (Colson and Hudson, 1976), Georgia (Golden and Price, 1976), Tennessee (Wibben, 1976), and Illinois (Curtis, 1977), and a sample of 99 streamflow stations was selected to give an approx imately uniform spatial configuration of station locations over the six-state study area (fig. 2). Calibrations were rerun for sites in Missouri, Tennessee, Alabama, and Georgia, to conform to a restrictive set of calibration guide lines, as follows, assign RR 9 EVC, DRN, and TP/TC (table 1). Recalibration results were essentially the same as those initially reported. The sample was divided into two sets: 50 stations for development of synthetic annual- flood data, and the remainder withheld, to be included in the development and evaluation of a map-model estimating procedure.
Synthesis of Annual Floods
At each of the 36 long-term rainfall sites shown in figure 2, and described in table 2, synthetic data were developed to relate rainfall-runoff model estimates of T-year floods to the parameters of the model. This was accomplished by generating a sample of 50 synthetic, annual-flood series using data from each rainfall site. The synthetic data sets were developed using calibrated model parameters for a representative sample of 50 basins shown in figure 2, and described in table 3. Replicate synthesis was performed at each rainfall site using the same sample of fitted model parameters, resulting in a total of 1,800 synthetic, annual-flood series (50 parameter sets by 36 rainfall records) that are representative of model applications in small rural basins.
Additional flood series were generated in an analogous manner to study the effects of "urban development" on synthetic flood characteristics. Ten levels of imperviousness (5, 10, ..., 50 percent) were assigned to each model/rain-gage application, resulting in a total of 18,000 (50 by 36 model/rain-gage applications by 10 levels of imperviousness) synthetic, annual- flood series with impervious effects. The generation of rainfall excess from impervious surfaces was modeled by a simple threshold concept that required the retention of 0.05 inch of storm rainfall before the surface becomes 100 percent effective in producing runoff. Retention capacity was allowed to recover by evaporation during periods of no rainfall.
IOW
AD
31
( C
hica
go
V.D
39
(*
T36
,-T
>3
2
ILL
.T
35
I
'T2
1
. D
36
I -
_-.
.'
0ft
on
a
JU T24
D40
[T32
D3
4\D
33
T3
4
i S
prin
gfie
ld'
IND
.
EX
PLA
NA
TIO
N
Dub
uque
, Io
wa L
ong
term
reco
rdin
g r
ain
fall
site
,
see
table
2.
D01
Map
cod
e fo
r st
rea
mflo
w s
tation,
see
tab
le 3
.
T01
~
M
ap c
ode
for
stre
am
flow
sta
tion
, /
see
table
3.
,/N
.J
/
D23
I D
42M
O.
T39
D43
"S
T44
T41
> °J
i-
Lou
isvi
lle
KY
.
Atl
anti
c C
ity
'\Vut
hevi
lle
Ric
hmon
d,
'tync
hbur
g
AR
K.
/
D44
M
D29
T27
£ M
emph
is
D3
°Cha
ttan
ooga
^
Utt
le R
ock
/ -
-D4
5- -
-j -
y D
Q9
/ D
4fi
T
«
AL
A-
\ T1
° D
08
X
D
48
T
45
I D
16
\D10
D01\
f T
46
^15
T16D
19
' -
/ T
17
N.C
.
^Cha
rlot
te
"--
--" N
_
_
Gre
envi
lte"
^T1
1 T
04T
V *
cr
Atl
anta
TO
rN
S.C
.M
ISS
."",
' D
20_» T
13."
\ D
02
^Aug
usta
T46
D4
9
D1
7ss
. ;
020
9 1.1
3 \
-
»x-
T21
,T18
B
irni
mgh
anu
QA
> \
-
D
13\
\D
07
^Mac
on D
04
' *
TO
zS
avan
nah
DO
S T
03
I T
47
I T
12
D13
A S
hrev
epor
t U
ickB
burg
D22
| M
ontg
omer
/M 3
TOfi\
,
t*,
.^M
erid
ian*
TH
9 D
14
/S^M
erid
ian
TH 9 i
T15
0 50
10
0 15
0 20
0 M
ILE
S
0 50
15
0 2
50
KIL
OM
ETE
RS
The log-Pearson Type III distribution, fit by the method of moments, was used to quantify synthetic T-year flood estimates for each model/rain-gage application.
Formulation of Multiple-Regression Model
The rainfall-runoff model simulates a flood hydrograph and associated peak discharge rate by routing a volume of runoff (pattern of rainfall excess) with a model of an instantaneous unit hydrograph, IUH. With this two-phase operation as a guide, two factors were defined an infiltration factor, F 9 and a hydrograph shape factor, L, that characterize the separate effects of model parameters on the volume of runoff and the shape of the IUH. Using these two factors as independent variables, a multiple-regression model was formulated to give a generalized definition of synthetic T-year floods (dependent variable) for each rainfall record, as follows,
~>
a.L. HF . 2i , (2) * 3 3
where
A
q. . = predicted value of standardized, synthetic discharge in cubic feet per second per square mile for the i-th recurrence interval, i = 2-, 25-, and ^00-years, for the j-th model application, j = 1 to 50 (drainage area is a scalar multiplier in rainfall-runoff model computations; therefore, the results were "standardized" by dividing by the drainage area),
L. = lag of IUH9 see table 3,e7
F. = infiltration factor, see table 3,
a. - regression constant for i-th recurrence interval, and*u
b-.jbp. = regression coefficients for i-th recurrence interval.
Lag time has been successfully used as a hydrograph shape factor in observed flood-frequency studies by Carter (1961), Martens (1968), and Anderson (1970), and was selected for use in the analysis of synthetic flood- frequency characteristics. The lag, L, of the rainfall-runoff model routing component is the expected value of the distribution of arrival times of the ordinates of the IUH. According to Kraijenhoff van de Leur (1966), the lag of the rainfall-runoff model IUH is a linear combination of the
reservoir recession coefficient, KSW, and the time base, TC, of the isosceles translation hydrograph
L = KSW + 0.5TC, (3)
where
L = lag of instantaneous unit-hydrograph, in hours, KSW = linear reservoir recession coefficient, in hours, and TC = time base of isosceles translation hydrograph, in hours.
An infiltration factor, F, was formulated to characterize the interactive effects of infiltration component parameters on the volume of storm runoff. The factor is defined as the infiltration capacity associated with an anteced ent soil-moisture condition of 85 percent of field capacity, BSM/BMSM = 0.85, and accumulated storm infiltration of two inches, SMS =» 2. These conditioning values of SMS and BMS/BMSM were estimated by analysis of the value and sensi tivity of the standard error of estimate for trial regressions using various combinations of these values. The form of infiltration equation used in rainfall-runoff model computations is
FP = KSAT[1.0 + (PSP/SMS)*(RGF'(1.0 - BMS/BMSM) + (BSM/BMSM))], (4)
and substitution of the conditioning values of SMS = 2, and BMS/BMSM = 0.85, yields the formulation for F as
F = KSAT[1.0 + 0.5-PSP*(0.15 RGF + 0.85)] (5)
Regression Analysis of Synthetic T-Year Floods Rural Model Applications
For each rainfall site, the synthetic T-year floods representative of the rural model applications (sample of 50) were related to rainfall-runoff model parameters by use of equation 2. The magnitude of the regression constants and coefficients will reflect the influence of climate as it is imbedded in the data used for synthesis. The rainfall-runoff model did not change from rain fall site to rainfall site, nor did the sample of model parameters. The only thing that influences the variability in output, from rainfall site to rain fall site, is the difference in input data a reflection of climate.
The multiple-regression analyses show an increase in the accuracy of the relations with increasing recurrence interval, and also in a north-to-south direction (table 4). This is due to a decrease in the variability, and an increase in the average level of modeled antecedent soil-moisture associated with increasing recurrence interval, and with higher rainfall in the South. The regression model does not explicitly include the influence of the varia bility of the parameter BMSM 9 and to a certain degree, the affect of its variability on synthesis results is attenuated by high rainfall, both north- to-south and with increasing recurrence interval.
The regression coefficient, b^ , which describes the influence of hydro- graph shape on T-year floods, shows low variability both geographically and with increasing recurrence interval. This is a reflection of the linearity of the rainfall-runoff model routing component double the volume, double the peak. The regression constant, a, and coefficient, £><>, show strong geographic variability, and are highly related as shown in figure 3. This empirical relationship indicates that the rainfall-runoff model is "well behaved" in the sense that it performs in a systematic manner when different rainfall records are used to synthesize annual floods. More importantly, the relationship suggests that the regression model could be modified by expressing the coeffi cient bg a function of a£, and thereby eliminate the need to define the
t*
geographic variability of £><>.. That is,t*
'9 J v*-* / i c"-'y ^^ P ' ^ 'o Is Is
Is
where the parameters y = 0.41, and $ = -1.39, describe the line approximating the relationship indicated in figure 3. In addition, the low variability ofthe regression coefficient, frj., (coefficient of variation = 0.13) suggests
i, _that it could be assigned its mean value, b^ = -0.69, with little loss of accuracy.
A second, "constrained" regression model was formulated as
-0.69 f(a.) q. . = a.L. F ^ , (7)
and the single coefficient, <X£, was determined for each recurrence interval, to minimize the sum of the squares of the difference in the logarithms of
50 2 flow by a direct search procedure; that is, Mi.n Z (log q. . - log q. .) ."^ "^
EXPLANATION
O 2- year recurrence interval
9 25- year recurrence interval
A 100-year recurrence interval
-0.7
100 400 1000
REGRESSION CONSTANT, a t
Figure 3. Relation between regression constant, a^, and regression coefficient,
10
The results shown in table 5 indicate that the constrained, single-coefficient regression model is a reasonable substitute for the initial multiple- regression model, with only slightly higher standard errors of estimate and similar trends in the accuracy of the relations as a function of recurrence interval and geographic location.
Effects of Imperviousness on Synthetc Floods
The unit-hydrograph concept of linearity, that relates flow rate to volume of runoff (double the volume, double the peak), was used to formulate a model to quantify the effect of the increased volume of runoff from impervi ous areas on synthetic T-year floods. If we neglect the difference in the time patterns of rainfall excess generated from pervious and impervious con tributing areas, then
PI ^ VI * 7(2-1)~ ~
where
PT = peak with impervious area contribution,
P = peak without impervious area contribution,
VT = volume runoff with impervious area contribution,
V = volume runoff without impervious area contribution,
I = proportion of drainage area with impervious surface,
R = volume of storm rainfall.
The ratio on the right-hand side of the approximation is analogous to the "coefficient of imperviousness", K, introduced by Carter (1961) to describe the variability of floods from urban watersheds. By rearranging terms, then
PT = P[l + I (|- 2)]. (9)
If we consider that the regression relation without impervious effects indi cates that
(10)
11
and that
V « , (11)
and specify
d <* R, (12)
then a general regression relation for both rural and impervious area model applications may be defined as
q. . ~ g.£."g' ggF/(Vl3.0 + J.( * , - 1.0)]. (13) ^,J i 3 3 3 p f(&£
3
The unknown coefficient, d^ 9 can be determined by a least-squares fit, just as O£ was determined in the analysis of the rural model applications. That is,
500 2 M-in I. (log q. . - log q. .) , (14)
by a direct search procedure. A summary of the results of the direct search determinations of the coefficient, d^ 9 is shown in table 6. The standard errors for the 2-year flood relations are consistently lower than those for rural-area model relations. This is because there is substantially less vari ability in the volumes of runoff and peak flow rates associated with the impervious-area model applications there are minimum percentages of runoff (5, 10, 15, ...50 percent) from every storm event. Therefore, the adequacy of the infiltration factor, F, to characterize runoff volume is not as critical in the regression model formulation for impervious-area applications because there is a "platform" of runoff from every storm event.
Single-Coefficient Relation for Synthetic T-Year Floods
Similarly, as in the case of the initial rural-regression results, there is a strong relationship between the magnitudes of the regression parameters, in this case, a^ and d^. This relationship, as shown in figure 4, indicates that two line segments are required to approximate the trend. Using these two approximating relations, equation 13 can be modified to yield a
12
10
CJ
oCJ
EXPLANATION
O 2-year recurrence interval
4) 25- year recurrence interval
100- year recurrence interval
40 100 400
COEFFICIENT, a i
1000
FIGURE 4. Relationship between the coefficients a-, and d«.
13
single-coefficient relation, a generalized definition, that describes both rural- and impervious-area model applications
-0.69 /(a.) g(a.)q. = a.L F * [1.0 + I.( * - 1.0)] (15) ^ ^ J J J T Jv";
where
/(a.) = 0.4JZ Z<?# a. - 1.39,
and
-0 71 a, for a. < 300, andi 1* ~
1 a.~°- 43 for a. >
The effectiveness of equation 15 to explain the variability of synthetic T-year floods is summarized in table 7. These results show that the general ized definition is a reasonably accurate relation that describes the results of synthesis for both rural- and impervious-area model applications. The average accuracy of the defining relation is somewhat better for the 25- and 100-year recurrence interval floods than that for 2-year floods, particularly in the North.
Geographic Variability of Climatic Factor, C
The site-to-site variability in the magnitude of the regression coeffi cient, a.£, in equation 15, is interpreted as reflecting the spatially varying influence of local climatic factors, C{,, on synthetic flood potential. Con tour maps depicting the geographic variability of the climatic factor, Cj, t that is, a£, are shown in figures 5, 6, and 7. Estimates of the climatic factor, used in conjunction with the generalized definition of synthetic flood potential, equation 15, offer a means of integrating long-term rainfall infor mation into a procedure to improve T-year flood estimates at rainfall-runoff model calibration sites. The remainder of this report develops and evaluates such a procedure.
14
M2
2
j V
alu
e o
f clim
atic f
act
or,
0 50
100150200
MIL
ES
I
i1
i '.
0 50
15
0 25
0 K
ILO
ME
TER
S
FIG
UR
E 5
. C
onto
ur
map
show
ing t
he g
eogra
phic
variatio
n i
n th
e 2
ye
ar
recu
rrence
inte
rval
clim
atic f
acto
r,C
2-
I V
alu
e o
f clim
atic f
act
or,
C
A8
44
0 50
100
15
02
00
M
ILE
SI
i1
i 'i r
-050
150
250
KIL
OM
ETE
RS
FIG
UR
E
6. C
onto
ur
map
sh
ow
ing
th
e g
eogra
phic
va
ria
tion
in
the
25
year
recu
rrence
in
terv
al
clim
atic f
acto
r.r_
_.
for
Dub
uque
, Io
wa.
I
TT
i^M
moL-^
^0
50
100
15
02
00
M
ILE
SI
H-r-H
r H
n
Rn
15
0
25
0
KIL
OM
ET
ER
S
FIG
UR
E 1. C
onto
ur
map
show
ing t
he
geogra
phic
variatio
n i
n th
e 10
0 y
ea
r re
curr
ence
inte
rva
l clim
atic f
act
or,
C-\
'
MAP-MODEL ESTIMATES OF T-YEAR FLOODS AT CALIBRATION SITES
Map-model estimates of the 2-, 25-, and 100-year recurrence interval floods can be computed for rainfall-runoff model calibration sites as follows:
1. Determine values of the climatic factors, £3 , C^ , and site location using figures 5, 6, and 7.
2. Compute model lag, L 9 using equation 3.3. Compute the infiltration factor, F 9 using equation 5.4. Compute the standardized discharge as,
f°r tne
q.=C.L-0.69 f(C.)
F ' g(C.) L*
- 1.0)1, (16)
where
f(C.) = 0.41 log C. - 1.39,L* Is
162C. °' 71 C. < 300, and ^ ^ '
32.9C ~°' 43 c. > 300. ^ ^
5. Compute map-model estimates of the T-year floods as,
(17)
where
A = drainage area in square miles
If it is assumed that the log-Pearson Type III distribution is the under lying population defining the frequency of map-model flood estimates, then_ estimates of the distribution parameters, that is, estimates of the mean, X, the standard deviation, 5, and the skew coefficient, (7, can be computed by using the values for the 2-, 25-, and 100-year flood magnitudes, and table
18
look-up procedures involving percentage points of the distribution as tabula ted by Barter (1969), as follows:
1. Use the values for 5g» Q25* an^ $100 as Previously determined, to compute the factor R, using
R =(log - log
Q200 - log(18)
and determine the skew coefficient, £, from the following table relating R and G.
3.02.92.82.72.62.52.4 2.32.2 2.1
0.99712.99590.99431.99220.98946.98600.98170.97651.97032.96351
-2.0-1.9-1.8-1.7-1.6-1.5-1.4-1.3-1.2-1.1
0.95495.94582.93585.92516.91392.90229.89044.87853.86672.85509
-1.0-0.9- .8- .7- .6- .5- .4- .3- .2- .1
0.84377.83279.82222.81206,.80235.79306.78419.77573.76766.75993
0.0.1.2.3.4.5.6.7.8.9
0.75255.74547.73870.73220.72594.71991.71408.70844.70298.69767
1.01.11.21.31.41.51.61.71.81.9
0.69250.68745.68252.67768.67293.66825.66364.65908.65457.65009
2.02.12.22.32.42.52.62.72.82.93.0
0.64563.64120.63678.63236.62794.62352.61908.61464.61017.60568.60117
The above table may be extended using the relation
R =(K100 ~ V
(19)
where Kj, is the Pearson Type III deviate (Water Resources Council, 1976, appendix 3.)
2. Compute the standard deviation, 5, using
S =DK (20)
19
where DK is determined as a function of skew coefficient, G 9 from the following table.
DK DK DK DK DK
The above table may be extended using the relation
DK
3.0
2.92.8 2.7 2.62.52. A2.32.22.1
0.27031.29845.32874.36125.39604.43314.47253.51423.55815.60423
-2.0-1.9-1.8-1.7-1.6-1.5-1.4-1.3-1.2-1.1
0.65233.70229.75393.80705.86144.91686.97307
1.029881.087081.14446
-1.0- .9- .8- .7- .6- .5- .4- .3- .2- .1
1.201871.259131.316141.372741.428851.484381.539231.593361.646741.69918
0.0.1.2.3.4.5.6.7.8.9
1.750691.801241.850811.899421.946901.993242.038412.082382.125102.16655
1.01.11.21.31.41.51.61.71.81.9
2.206662.245412.282752.318632.353032.385872.417152.446812.474822.50113
2.02.12.22.32.42.52.62.72.82.93.0
2.525732.548582.569662.588942.606432.622092.635952.648002.658232.666672.67334
DK = (K25 (21)
3. Compute the mean, J, using,
X = log - K2S (22)
where K is a function of skew, G, as shown below.Ci
G
-3.0 2.9 2.82.7 2.62.52.4 2.32.2 2.1
K2
0.39554.38991.38353.37640.36852.35992.35062.34063.32999.31872
G
-2.0-1.9-1.8-1.7-1.6-1.5-1.4-1.3-1.2-1.1
K2
0.30685.P9443.28150.26808.25422.23996.22535.21040.19517.17968
G
-1.0-0.9- .8- .7- .6- .5- .4- .3- .2- .1
K2
0.16397.14807.13199.11578.09945.08302.06651.04993.03325.01662
G
0.0.1.2.3.4.5.6.7.8.9
K2
o.ooooo-.01662-.03325-.04993-.06651-.08302-.09945-.11578-.13199-.14807
G
1.01.11.21.31.41.51.61 .71.81.9
K2
-0.16397-.17968-.19517-.21040-.22535-.23996-.25422-.26808-.28150-.29443
G
2.02.12.22.32.42.52.62.72.82.93.0
K2
-0.30685-.31872-. 32999-.34063-.35062-.35992-.36852-.37640-.38353-.38991-.39554
Using these values for the mean, standard deviation, and skew, additional map-model T-year flood estimates can be computed for any desired recurrence interval using the relation
Q. = antilog (X + K.S). (23)
20
Comparison of Observed and Map-Model Flood Estimates
Observed and map-model estimates of T-year floods were computed for 98 rainfall-runoff model calibration sites shown in figure 2, and summarized in table 3. The sample includes 49 of the 50 basins initially selected for use in developing the synthetic flood relations, plus an additional selection of 49 sites to add "strength in numbers"; thus, it was hoped to yield a more meaningful sample for use in comparing observed and map-model flood estimates, Observed T-year flood estimates, Q^ 9 were computed using procedures recom-^ mended by the Water Resources Council (1976). Map-model flood^ estimates, Q^ 9 and associated log-Pearson Type III distribution statistics, X 9 S 9 and G were computed using the methods previously outlined.
*N* A.
Scatter diagrams of observed, Q^ 9 versus map-model, Q^ 9 flood estimates are shown in figure 8. The plots indicate that map-model estimates of the 25- and 100-year floods are generally lower than observed estimates. A com parison of the average values of distribution statistics and T-year floods, in 1og-,Q units, is shown below.
Average Observed Map-model
X 2.497 2.499S 0.298 0.264G -0.109 -0.261
Iog10 Q2 2.502 2.512Iog10 Q25 3.004 2.929
3 - 161 3 ' 049
If it is assumed that on the average, the observed estimates comprise an unbiased time-sample, that is, for each recurrence interval
98 , 7 987 «J ' W *
J -L 3 -
then the map-model estimates are apparently biased and should be adjusted to remove the discrepancy in average values shown above. Thus, the generalized definition was modified such that
"unbiased" <7. * £.<?., (25) n9
21
10.0
0010
.000
K3
K3
g
10.0
00
o
o LLJ
CO <r u CO D
O1,
000
CO
LL
J
Q O o Q LLJ
>
100
EX
PLA
NA
TIO
N
O 2
-ye
ar
recu
rrence
in
terv
al
25 -
year
recu
rrence
in
terv
al
A 1
00
-ye
ar
recu
rre
nce
in
terv
al
MA
P-M
OD
EL
FLO
OD
ES
TIM
AT
E,Q
, IN
CU
BIC
FEE
T PE
R S
EC
ON
D
FIG
UR
E 8
. S
catt
er
dia
gra
ms
of
ob
serv
ed
and
map-m
odel
T-y
ea
r flo
od e
stim
ate
s.
where
the average map-model bias, B^ = 0.98, B^ R 1.19, and B = 1.29
ACCURACY AND WEIGHTING OF OBSERVED AND MAP-MODEL T-YEAR FLOODS
We have two essentially independent estimates: the map-model estimate
Q. ., and the observed estimate Q. ., of the true but unknown T-year floodi»d ^»JQ. .. We seek to appraise the accuracy, in some average sense, of the map-^ jj
model estimating procedure, equation 25, and to develop a method of weighting observed and map-model T-year flood estimates. The logarithmic transform of the map-model estimate of the true flood can be written as
log q. . = log Q. . + a. ., (26)
where a. . reflects a lumping of error terms including:1 13
(1) error in the rainfall-runoff model (infiltration, routing, etc.),(2) error in the calibrated model parameters (KSW, KSAT , etc.),(3) error in the generalized synthetic flood relation, equation 15,(4) error in the maps depicting the influence of climate, C^ 9 on
synthetic T-year floods, including resolution, time- samp ling, and measurement errors, and
(5) error in determination and applicability of the average bias coefficients, B., equation 25.
(s
If we assume that the expected value of a. . has mean zero, that is,* *- >Jlog Q' .is unbiased, then the variance of a. . isy
7ar[a. .] = E[(log Q. . - log Q, .)]. (27)^ 9 J I* »«/ ^ > I/
Similarly, the logarithmic transform of the observed estimate of the true flood is given by
log Q. . = log Q. . + e . .. (28)
23
If we ignore errors In measurement of discharge and errors in assumptions of the underlying population distribution, then the error term e. ., is solely a
function of record length, that is, time-sampling error. If we assume that the expected value of e . . has mean zero, that is, Q. .is unbiased, then the variance of e . .is ^J
Var[e. .] = E[(log Q. . - log Q, .)] (29)
Although both the variance of a. . and e. . may vary from site-to-site,'Z'jJ 'Z'»J
we seek an overall, average measure of the error variance of the map-model estimating procedure, such as
__7or[a.J = Z El (log Q. . - log Q. .)"]; (30)
in terms of an overall, average measure of the time-sampling error variance of the observed flood estimates, such as
7 M ~ 2 ] = 7? £ E[(log Q. . - log Q. .) ]. (31)
y? J n/J * ^ t-/ T yfl x» lJ w >9 X7 **
A similar average measure of adequacy is found in regression analysis, where the standard error of estimate is taken as a measure of accuracy of regression estimates, even though the variance of any individual estimate is not constant for all values of the independent variables.
One method of obtaining an estimate of Vap]a.] is as follows. The^ ^
squared differences in log Q. ., and log Q. - can be, written as"Z-»J ^, j
24
(log Q. . - log Q. .) 2 = (log Q. . - log Q. . + log Qi . - log ^ ) ,^>«7 ^>t/ ^»V ^ »«/ 'V
- log Q.
^ ̂ - log QtJ. (32)
If the deviations (log' 5- - - log Q. .) and (^ Q. . - log Q. .) are con-t-»J t-»J t-»t7 ^>t/
sidered independent in a statistical sense, then by taking expectations and rearranging terms
or
E[(log Qi9J - log 2 )] = E[(log ^ - log j
. - log Q. -) 2 ], (33)
7ar[a. .] = Var{n^ .] - 7ar[e^. .], (34)
where
Var[n. .] = ^I(Zo^ 5- - log Q. .)]. (35)t- J J i- >J ^ >t/
According to Hardison (1971), 7ar[e. .] can be estimated by R . .£./#., that j ^>J ^>J J J
. .] = R2 . .S2 ./N., (36) - '
2 where R . . is a function of recurrence interval and skew coefficient (HardiW 2
son, 1971, table 2) , S .is the sample estimate of the variance of the
25
logarithms of observed annual peaks, and N. is the number of observed annualJ *
peaks. The quantities Var[n. .] can be directly estimated because both Q. .t t j 1,3and Q. . are available. Thus, averaging over all sites gives an estimate of
the average map-model error variance, Far [a.] asis
VMM. = SD. - VT., (37)Is Is 1s
where
- M Z J =
- ~ 2 SD. = Z (log Q. . - log Q. .r, (38)
and
1 M 2 2VT. = ~ Z E6 . .S . IN .. (39)33
The averaging factor, M-l , used in the computation of the mean squared loga
rithmic deviation, S£> . , accounts for a lost degree of freedom associated withIs
the determination of the average bias coefficients, B.. A word of caution isis
necessary when using this method of decomposing variance components. Because
££>. and VT. are computed independently of one another, negative values of VMM.is is is
may result, which are not meaningful.
The preceding formulaton is similar to that developed by Hardison, (1971, equation 3), in his study of prediction error of regression estimates of streamflow characteristics:
= VR " (1 ' p)V (40)
where
V' = variance due to space-sampling error; same as model error,oVp - variance of the regression, the square of the standard error of esti
mate of a regression,
p = average interstation correlation coefficient, and
F = average variance of the time-sampling error at the stations used in the regression.
26
The analogy between the two formulations is as follows: VMM is analogous to
7C , SD is analogous to FD and VT is analogous to 7 . The influence of theu A -t
average interstation correlation, p, is not accounted for in equation 37, but its absence is thought to be of minor importance in the present analysis (p - 0.06 for the 98-station sample).
^\ *v
If it is assumed that the two estimates Q. ., and Q. ., are independent* »J >>J
and unbiased, then a weighted estimate, WQ. ., can be formed as
WQ. . « Wifflf. . £. . + WO. .-Q. ., (41) -
where the map-model weighting factor is
. .S2 ./N.(42)
. J3 ./N. + VMM.) T-,3 C J ^
and the observed weighting factor is
W. . = 1 - WM. .. (43)
The variance of the weighted estimate is less than the variance of either the map-model or the observed estimate, and is given by
VMM.'P . .S ,/N. VW. . = ^ \^ 'I J . (44)
(VMM. + P . .5 ./tf.)J 3
27
The average value of the error variance of the weighted estimates may be expressed as
7~ £ VW.
Equations 37, 38, 39, and 45 were used to compute estimates of the aver
age values of the map-model error variance, VMM., the time-sampling variance,_ i>VT. 9 and the error variance of the weighted estimates, VW., using observed and
Is Is
map-model flood data for the test sample of 98 rural-area calibration sites. The trends in the values of these variance components are shown as a function of recurrence interval in figure 9. The rapid decrease in the magnitude of the average map-model error variance, from the 1.25-year to the 5-year recur rence interval, is a reflection of the accuracy of the synthetic T-year flood relation, as a function of recurrence interval, and the general accuracy characteristics of the rainfall-runoff model in the reproduction of small- magnitude floods.
The synthetic T-year flood relation is less accurate at low recurrence intervals because of the absence of the parameter, BMSM , and because the influence of modeled antecedent soil-moisture conditions, BMS , and SMS , are not explicitly included, nor can they be explicitly included, in the defining relation. SMS and BMS are model variables, not model parameters, and their influence on synthesis results can only be approximated in an average sense, that is, by the "conditioning" values, as previously determined. Modeled antecedent conditions, the influence of BMSM, and the adequacy of the infil tration factor, F 9 are less important in explaining synthetic results for the higher recurrence interval floods, because they are attenuated.
Similarly, the influence of real-world, antecedent soil-moisture condi tions are only approximately modeled by the rainfall-runoff model; they are very important in explaining the variability of flood magnitudes from lower magnitude storms, and resulting generally lower magnitude floods. The rainfall-runoff model is more accurate in reproducing extreme floods than minor ones. Extreme floods from small basins are extreme for several reasons, but high rainfall intensity/duration dominates. In general, infiltration losses become a less significant part of total rainfall for extreme rainfall events a threshold effect. The inadequacies in modeled soil-moisture condi tions and in the infiltration component tend to be attenuated for the larger events. Storm rainfall assumes a dominant role in determining the output, both in nature and in the rainfall-runoff model abstraction of it, as well as in the regression model abstraction of rainfall-runoff model.
28
0.04
CO
EXPLANATION
Average map-model error variance , VMM
O Average observed time-sampling error variance . VT
A Average error variance of weighted estimate ,VW
0.001.25 5 10 25
RECURRENCE INTERVAL, IN YEARS
100
FIGURE 9. -- Trends in error-variance components as a function of recurrence interval.
29
The average map-model error variance decreases to a minimum value of about 0.011 squared log units at the 10-year recurrence interval; then reverses its trend and increases with increasing recurrence interval a desirable out come. Without this reversal, the ascribed value of the map-model estimating procedure, in terms of the concept of equivalent-year accuracy (Carter and Benson, 1970), would increase without limit. Logic tells us that there is an upper limit to the amount of information that can be extracted from long-term rainfall records, in this case about 65 years, because that is the average length of record used in the development of the map-model estimating procedure. Obviously, this is an unattainable upper limit, because of the many sources of error inherent in the map-model estimating procedure, including errors in the rainfall-runoff model, the generalized definitions, and the maps depicting the geographic variability of the inferred effect of climate on synthetic floods.
The trend in the average time-sampling error variance of the observed T-year flood estimates, with a minimum near the 2-year recurrence interval, is a completely predictable statistical outcome. This is because the 2-year flood is less affected by time-sampling error than are the flood levels asso ciated with the tails of the distribution, where the error in sample estimates of the standard deviation and skew coefficient is more influential.
The average error variance of the weighted flood estimates, VW. 9 showsuthe worth of the map-model estimating procedure in relation to observed data estimates that have a harmonic-mean record length of 13.2 years. Note partic
ularly the "flatness" in the trend of VW. as a function of recurrence interval,"i*and that the average accuracy of the weighted estimates of the 100-year floods is equivalent to that of the observed data estimates of the 2-year floods. The average accuracy of the weighted T-year flood estimates can be appraised in terms of equivalent-length of observed record using the relation
VT.m. = NG ̂ r (46)
* VW.
where
NE. - average equivalent record length of weighted estimates, in years,Is
NG = harmonic-mean record length of observed annual flood records (13.2 years).
The following table shows estimates of the equivalent record lengths of the weighted flood estimates at selected recurrence intervals.
30
Recurrence interval Equivalent record length(years) (years)
1.25 192 205 28
10 3425 4050 43
100 44
These results show a pronounced nonlinearity in equivalent record length as a function of recurrence interval, and also indicate an upper limit to the amount of peak-flow information attainable by the map-model estimating procedure a gain of about 30 years of record for estimating the higher recurrence interval floods (the 50- and 100-year floods), as compared to a gain of about 6 years of record for estimating the lower recurrence interval floods (the 1.25- and 2-year floods).
SUMMARY AND CONCLUSIONS
Annual flood series were synthesized for a wide variety of modeled, hydrologic conditions using data from each of 36 long-term recording rainfall sites. A single-coefficient regression relation was developed for each rain fall site to give a generalized definition of synthetic T-year floods as a function of the rainfall-runoff model parameters. The accuracy of the general ized definition increases with an increase in recurrence interval, with an increase in percent impervious area, and in a north-to-south direction. These trends in accuracy are a ramification of the decreasing influence of the vari ability in infiltration component parameters (KSAT 9 PSP, RGF, and BMSM) on the increased volumes of runoff and peak discharge rates associated with increasing recurrence interval, increasing impervious area, and also with higher rainfall in the South. The extreme floods are extreme for several reasons, but high rainfall intensity/duration is a necessary condition. For the extreme events, modeled infiltration losses become a less significant part of total rainfall, and the effect of the site-to-site variability in infiltration component param eters is attenuated, as it should be. The site-to-site variability in the routing component parameters, those parameters that define the hydrograph shape factor L, assume a dominant role in explaining the reduced variability of extreme flood events. A general maxim for modeling the results of synthesis is "...the higher the variability, the more model data input required to explain the variability." The degree of variability, and thus the ability of the generalized definition to explain that variability, is inversely related to the magnitude of the flood event it takes less model to explain the reduced variability in extreme floods, and more model to explain the higher variability in small magnitude events.
31
The site-to-site variability in the magnitude of the regression coeffi cient, a, that characterizes the generalized T-year flood relation, is a reflection of the spatially varying influence of climate, C 9 on the results of synthesis. Maps depicting the geographic variation in the magnitude of C were prepared and used in conjunction with fitted rainfall-runoff model parameters and the generalized synthetic flood relation to develop map-model, T-year flood estimates for 98 small-basin calibration sites. Comparisons of these T-year flood estimates with those based on observed annual floods show that the map-model estimating procedure is biased the map-model estimates are generally lower than the observed estimates for return periods greater than the 2-year recurrence interval. This tendency for underestimation of the higher recurrence interval events may be attributed to several factors, including:
1. A loss of variance associated with a model smoothing effect, as described by Matalas and Jacobs (1964), and Kirby (1975);
2. The effect of unmodeled, real-world nonlinearities in the transfor mation of rainfall excess to discharge hydrograph (routing) a limitation of the unit-hydrograph concept as used in the rainfall- runoff model;
3. Incorrectly modeled nonlinearities in the synthesis of rainfallexcess (volume of runoff), due to inadequacies in either anteced- dent soil-moisture accounting or infiltration computations;
4. Sampling errors in the long-term rainfall data used for synthesis of annual floods;
5. The use of an average daily pattern of potential evapotranspiration.
Regardless of the particular cause or causes for the tendency to underestimate the higher recurrence interval floods, the bias can be removed in an average sense by using an average adjustment factor, that is, the bias coefficients,B..^
The average accuracy of unbiased, map-model flood estimates was appraised for the sample of 98 rural-area calibration sites located in a six-state study area. This appraisal shows that the accuracy of the estimates increases rapidly, with increasing recurrence interval up to about the 10-year return period, and then reverses its trend and decreases slowly. This pattern of error variance, when compared to that associated with observed flood esti mates, indicates that the map-model estimates are more accurate than the observed estimates beyond the 10-year recurrence interval (the harmonic-mean record length of the observed annual flood series is 13.2 years).
Improved T-year flood estimates were developed by computing a weighted average of observed and map-model estimates, and the accuracy of the improved estimates was appraised as a function of recurrence interval and in terms of the concept of equivalent-length of record. The trend in the accuracy of the improved estimates shows that the map-model estimating procedure yields an equivalent-length of observed record that ranges from a low of about 6 years for the 1.25-year flood, up to an ultimate, maximum level of about 30 years of data for estimating the 50- and 100-year recurrence interval floods.
32
REFERENCES
Anderson, D. G., 1970, Effects of urban development on floods in Northern Virginia: U.S. Geological Survey Water-Supply Paper 2001-C, 22 p.
Benson, M. A., 1962, Factors influencing the occurrence of floods in a humid region of diverse terrain: U.S. Geological Survey Water-Supply Paper 1580-B, 64 p.
____1964, Factors affecting the occurrence of floods in the Southwest: U.S. Geological Survey Water-Supply Paper 1580-D, 72 p.
Carrigan, P. H., Jr., 1973, Calibration of U.S. Geological Survey rainfall- runoff model for peak flow synthesis natural basins: U.S. Geological Survey Computer Contribution, 114 p. Available only from U.S. Department of Commerce, National Technical Information Service (NTIS), Springfield, Va. 22151, accession no. PB-226-217.
Carter, R. W., 1961, Magnitude and frequency of floods in suburban areas, in Short papers in the geological and hydrologic sciences: U.S. Geological Survey Professional Paper 424-B, p. B9-B14.
Carter, R. W., and Benson, M. A., 1970, Concepts for the design of streamflow data programs: U.S. Geological Survey Open-File Report, 33 p.
Clark, C. 0., 1945, Storage and the unit hydrograph: American Society of Civil Engineers Transactions, v. 110, p. 1419-1488.
Colson, B. E., and Hudson, J, W., 1976, Flood frequency of Mississippi streams U.S. Geological Survey Open-File Report, 34 p.
Curtis, G. W., 1977, Frequency analysis of Illinois floods using observed and synthetic streamflow records: U.S. Geological Survey Water-Resources Investigations 77-104, 32 p.
Dawdy, D. R., Lichty, R. W., and Bergmann, J. M., 1972, A rainfall-runoff model for estimation of flood peaks for small drainage basins: U.S. Geological Survey Professional Paper 506-B, 28 p.
Golden, H. G., and Price, M., 1976, Flood-frequency analysis for small natural streams in Georgia: U.S. Geological Survey Open-File Report 76-511, 77 p
Hardison, C. H., 1971, Prediction error of regression estimates of streamflow characteristics at ungaged sites: U.S. Geological Survey Professional Paper 750-C, p. C228-C236.
Harter, H. L., 1969, A new table of percentage points of the Pearson Type III distribution: Technometrics, v. 11, no. 1, p. 117-187.
Hauth, L. D., 1974, Model synthesis in frequency analysis of Missouri floods: U.S. Geological Survey Circular 708, 16 p.
Ktrby, W., 1975, Model smoothing effect diminishes simulated flood peak vari ances Cabs): American Geophysical Union Transactions, v. 56, no. 6, p. 361.
Kraijenhoff van de Leur, D. A., 1966, Runoff models with linear elements: in Committee for Hydrological Research T.N.O., Recent trends in hydrograph synthesis: Verslagen En Mededelingen, Commissie Voor Hydrologisch Onderzoek T.N.O. 13, The Hague, p. 31-62.
Martens, L. A., 1968, Flood inundation and effects of urbanization in metro politan Charlotte, North Carolina: U.S. Geological Survey Water-Supply Paper 1591-C, 60 p,
Matalas, N. C,, and Jacobs, B., 1964, A correlation procedure for augmenting hydrologic data: U.S. Geological Survey Professional Paper 434-E, 7 p.
33
McCain, J. F., 1974, Progress report on flood-frequency synthesis for small streams in Alabama: Alabama Highway Research, HPR70, 109 p.
Philip, J. R., 1954, An infiltration equation with physical significance: Soil Science Society of America Proceedings, v. 77, p. 153-157.
Thomas, D. M., and Benson, M. A., 1970, Generalization of streamflow character istics from drainage-basin characteristics: U.S. Geological Survey Water- Supply Paper 1975, 55 p.
Water Resources Council, 1976, Guidelines for determining flood-flow frequency: U.S. Water Resources Council Bulletin 17, 26 p.
Wibben, H. C., 1976, Application of the U.S. Geological Survey rainfall-runoff simulation model to improve flood-frequency estimates on small Tennessee streams: U.S. Geological Survey Water-Resources Investigations 76-120, 53 p.
34
Table 1. Model parameters and variables and their application In the modeling process
Parameter Variable Units Application
BMSM-
RR
EVC-
DRN-
BMS-
SMS-
FR
KSAT -
PSP- -
RGF-
KSW
TC
TP/TC-
SW-
Inches Soil-moisture storage at field capacity. Maximum value of base moisture storage variable, EMS.
0.85* Proportion of daily rainfall that infiltrates the soil.
0.65-0.75* Pan evaporation coefficient.
1.0* Drainage factor for redistribution of saturated moisture storage, SMS, to base (unsaturated) moisture storage, BMS, as a fraction of hydraulic conductivity, KSAT.
Inches Base (unsaturated) moisture storage in active soil column. Simulates antecedent moisture content over the range from wilting-point conditions, BMS=03 to field capacity,BMS=BMSM.
Inches "Saturated" moisture storage in wetted surface layer developed by infiltration of storm rainfall.
Inches per Infiltration capacity, a function of KSAT, PSP, RGF, BMSM, hour SMS, BMS (equation 4).
Inches perhour Hydraulic conductivity of "saturated" transmission zone.
Inches Combined effects of moisture deficit, as indexed by EMSt and capillary potential (suction) at the wetting front for BMS equal to field capacity, BMSM.
- Ratio of combined effects of moisture deficit, as indexed by BMS, and capillary potential (suction) at wetting front for BMS=0=w±lting point, to the value associated with field capacity conditions, PSP.
Hours Linear reservoir recession coefficient.
Minutes Time base (duration) of triangular translation hydrograph.
0.5* Ratio of time to peak of triangular translation hydrograph to duration of translation hydrograph, TC.
Inches Linear reservoir storage.
*The parameters RR and EVC are highly "interactive" and were constrained. RR was arbitrarily assigned the value of 0.85, and EVC values were computed as the factor required to scale available local average annual pan evaporation to equivalent values of average annual lake evaporation as estimated from figure 2, "Evaporation maps for the United States," Technical Paper No. 37, U.S. Department of Commerce, 1959. The parameters DRN and TP/TC have little influence on model results. DRN was arbitrarily assigned a value of 1.0, and the shape of an isosceles triangle assumed for the translation hydrograph.
35
Table 2. Long-term recording rainfall sites used in study
[NS, number of storm events; NWY, number of water years]
Rain-gage location
P'Hr-arni Til _____
Peoria, 111. Springfield 111. Atlantic City, N.J.
Baltimore, Md.
St. Loiuis, Mo.
Springfield, Mo. Cairo, 111. Wytheville, Va.
Nashville, Tenn.
Charlotte, N.C.
Chattanooga, Tenn.-
Greenville, S.C. Little Rock, Ark. Columbia, S.C.
Montgomery, Ala.
Jacksonville, Fla.-
New Orleans, La.
Period of record
1 qn^ 71 ___ _ _1 QO? 7 A ___ _____ ___ _ _ _________ .1 Qn^ 7A ___ . ______ .
1 QOO 1 L 1 Q1 Q RS __ _ __ _ _____ . _
1900 71 _ _____
1»QT_1QTJ 1 Qlc:_7n __ ___ ___________ ___ __ __
1 fiQfi 1 Q7fi _________ ____ __________ _____ __ ___
1893-97, 1899, 1902, 1905, 1907-65, 1969
1 8Q8_1 QfiQ ____________ ___ ______ ________________ __
1Qf)9_91 1Q91 11 1 Q17 1O ______ _ ... _ _ ___
1 QflS 1 1 1 Q1 ^ 70 __ __ ____ . _________
1 QD8-7A ____ __ _ __ ____ ___________ ______
1905-14, 1916-42, 1944-47, 1951-52, 1954-63, 1965-70
1 RQR_1 Q7fl ____ _________ ____
1 flQfl 1 Qfifi 18 1 Q9n_m 1QT7 71- _ .___. _ _________ _ .
1QO1-71 __ ___ _ ___ ____ ___ ___
1 Q1 Q_1O 1 QOQ 71 _ _ __ _ _____ _ _ _ __ ._
1 RQfl_1 Q7A__ _ ________ _ _ .____ _ ____ _ .__ _ ___________
1QD1 RA ___ _ __ __ __ _________ _ __ ________ ____
1 QQJ}_1 Q7Q __ __________ ___ _ ________ ____ ___ _____
1 QflA 7 1} __ ________ __ ___ ____ ___
1 QD9 7^ ___ __ ________ __ __ ______ __ _____
1 QDO 7"^ ___ ______ _________ ____ ________
1 Q1 T "^0 1 Qfi'', T) _ __ __ . - .____ _ . _ ___. _ .____
1898-1948, 1950-55, 1957-67 1 QQ7_1 01 o IQIR ^n _________ ________ _ ____ __________
i oofi (\~i ___ _ _______ _ __ _ _ _________
1 QflA 1'Q 1 QA 1 71 _ _. __ _
i One: 79 _ ____ ____ _____ _ _ ___ _ ___ __
1 Of)"} AS __ _ ___ __ ___ ___________ ___
1 Q1 9 79 _ _ _ ___ _ _ __ __ .____ __ __
NS
129 168 352 177 165
221 231 221 187 274
156 162 180 166 129
250 313 684 359 339
139 370 143 332 350
377 443 277 146
. 167
331 429 265 345 278 340
NYW*
69 73 70 7055
72 67 73 66 51
72 64 60 67 59
73 73 68 71 73
49 73 54 76 70
72 74 53 68 53
68 76 61 68 66 61
*Annual flood synthesis requires daily precipitation valuss during nonstorm moisture accounting periods and 5-minute rainfall intensities during storm events (fig. 1). A "storm event" is a sub jective definition and may comprise several hours of intense rainfall occurring within a period of several consecutive days. Dates of storm events were identfied by U.S. Geological Survey personnel by analysis of available precipitation summaries. The majority of the storm event data were coded from original recorder charts by the National Climatic Data Center, NOAA, Asheville, N.C. In some instances copies of original recorder charts were acquired and reduction made by U.S. Geological Survey personnel. Streamflow characteristics such as the annual peak discharge, mean annual flow, and so forth, are typically reported on a water-year basis. The water year, October 1 through September 30, is identified by the calendar year in which it ends.
36
Tabl
e 3. Streamflow st
atio
ns used in
st
udy
OJ
Map
code*, st
atio
n no., na
me,
loca
tion
, and
numb
er of ob
serv
ed an
nual
floods
D01
D02
D03
D04
DOS
D06
D07
DOS
D09
D10
Dll
D12
D13
D14
D15
B16
D17
D18
D19
D20
D21
D22
D23
D24
D25
3D26 D27
D28
D29
0219
1280
02
1919
30
0221
5245
0222
5210
02327350
0234
3700
02
3462
17
0238
1600
02384600
0238
8200
02399800
0241
0000
02
4134
00
02427300
02435400
02437800
02441220
02449400
02451750
0246
2600
02
4791
65
0248
2900
0334
4250
0343
5600
03461200
03486225
03535180
03539100
03597500
Mill
Shoal
Cree
k ne
ar Royston, Ga
. Bu
ffal
o Creek
near L
exington,
Ga.
Folsom Creek Tributary
near
Rochelle ,
Ga .
Hurr
ican
e Br
anch
ne
ar Wrightville,
Ochl
ocko
nee
Rive
r Tr
ibut
ary
near
Coolidge,
Ga.
Stevenson
Creek
near He
adla
nd,
Ala.
Cele
oth
Cree
k near M
anchester, Ga.
Fauc
ett
Creek
near
Ta
lkin
g Ro
ck,
Ga.
Mill
Cr
eek Tributary
near
Et
on,
Ga.
Stor
ey Mill
Cree
k ne
ar
Litt
le Terrapin Creek
near
Patt
erso
n Creek
near
Ce
ntra
l, Al
a.
Wedo
wee
Creek
abov
e We
dowe
e, Ala.
Prai
rie
Creek
near
Oak
Hill
, Al
a.
Clea
r Br
anch
near Tu
pelo
, Mi
ss.
Barn
Cr
eek
near
Kackelburg,
Ma.
Sand
Creek
Trib
utar
y ne
ar M
ayhe
w,M-ico ____ ________ _____ . __ ___________
Mosquito Br
anch
at
Benndala,
Miss.
Tallahogue Creek
Trib
utar
y ne
ar
Emba
ras
River
Trib
utar
y ne
ar
Sulp
hur
Fork R
iver
Tr
ibut
ary
near
Cosby
Creek
above
Cosb
y, Te
nn.
Powd
er Br
anch
near
Johnson
City,
Willow For
k ne
ar Ha
lls
Crossroads,
Byrd
Creek near Crossville,
Tenn.
Wartrace Cr
eek
at Be
ll Buckle,
12
12 12
10 11
15 7 10
11 10 8 24
13
14
20
14 10
16 9 14
21 10
20 8 17 4 9 9
79
Rainfall-runoff
mode
l pa
rame
ters
PSP
(in)
0.66
.65
1.03
1.91 .71
5.42
3.29
6.76
.87
1.45
1.95
7.85
3.35
2.19
2.
70
1.26
1.01
1.27
2.92
2.40
2.84
3.34
.78
5.75
4.08
6.36
5.72
4.
05
9 m
KSAT
(in/h)
0.10
5 .089
.098
.054
.029
.145
.191
.061
.097
.074
.070
.081
.091
.037
.049
.0
67
.034
.032
.087
.069
.036
.059
.027
.097
.073
.156
.085
.135 mn
EOF
25.8
14
.2
32.1
14.9
26.4
4.
1 9.
7 13
.8
4.9
5.0
15.5
16
.2
10.3
15.7
15.6
21.0
10.1
11.9
14.6
40
.1
7.9
10.0
24.6
17.6
22
.2
22.9 7.1
10.1 o
ft
BMSM
(i
n)
4.59
6.
08
7.58
3.10
4.04
9.
09
3.32
9.66
2.
30
9.06
3.90
6.
60
2.51
2.02
3.07
5.
65
2.70
2.
21
6.48
6.60
1.14
5.23
1.31
4.97
1.04
6.24
7.32
3.26
t Q'
)
KSW
(h)
0.57
3.53
3.12
7.27
5.22
3.10
2.
27
5.40
4.25
4.86
8.13
1.50
1.80
5.
00
2.44
8.00
1.50
1.
85
1.25
6.
00
1.95 .99
1.66 .65
2.30 .55
1.75
1.57
1 OS
TC
(h)
0.73
4.94
4.65
7.40
4.02
4.50
1.97
3.
00
5.50
3.74
6.50
1.
75
4.00
4.00
3.
73
5.00
1.25
5.
00
1.25
2.
00
4.05 .55
.40
.83
3.17
1.17
2.67
3.10
A
1 7
Para
mete
rs us
ed in map
-mod
el
esti
mati
ng procedure2
Area
(mi2)
0.32
5.80
1.44
3.53
1.81
12
.4
2.82
9.99
4.28
6.02
15.9
4.
95
6.50
9.73
.75
12.9 .4
4 11.7
1.64
5.70
.22
.12
.081
3.50
10.2 4.88
3.23
1.
10
ift
*
L (h)
0.94
6.00
5.45
11.0 7.23
5.
35
3.25
6.
90
7.00
6.73
11.4
2.38
3.80
7.
00
4.31
10.5 2.12
4.35
1.88
4.00
3.98
1.27
1.86
1.07
3.88
1.14
3.08
3.12
i n
F
(in/h)
0.27
.18
.39
.21
.079
.72
.92
.66
.16
.16
.29
1.12
.46
.17
.26
.24
.075
.087
.47
.64
.14
.29
.075
1.07
.6
9
1.48 .55
.78 ftaa
°2 175
185
250
225
300
310
225
185
185
210
230
285
250
325
260
265
280
310
260
280
450
300
145
180
135
110
135
170
onn
C2S 66
0 67
5
860
800
940
980
820
640
625
660
740
900
800
980
725
740
810
925
725
810
1,400
950
545
575
560
460
525
575
f.f\i\
°WO
900
950
1,25
0
1,17
5
1,325
1,400
1,150
925
900
950
1,025
1,250
1,12
5 1,450
1,000
1,00
0
1,15
0 1,325
1,00
0 1,150
2,000
1,350
750
800
900
700
850
850
me
Table 3.
Str
eamf
low
stat
ions
used in study Continued
00
Map
code
1, station
no.,
name,
location,
and
numb
er o
f observed annual floods
D30
D31
D32
D33
D34
D35
D36
D37
D38
D39
D41
D42
D43
D44
D45
D46
D47
D48
D49
D50
T01
T02
T03
T04
T05
0360
4070
0543
9550
05448050
0550
2120
05503000
0552
7050
05
5660
00
0558
7850
0559
9640
06816000
0690
2800
0691
0250
06
9217
40
0701
5500
07028940
0703
0365
07054200
0706
3200
0727
7550
0728
7140
0737
3550
02192300
0220
2910
0221
6610
0221
7400
0222
6150
Coon
Creek Tributary, Hohenwald,
Sout
h Branch Ki
shwa
ukee
River
Riser
Creek Tributary
near
Oak
Dale Branch near Emden, Mo
.
Prairie
Cree
k near Frankfort, 111.
East
Branch Panther
Creek
near
Gridley, 11
1.
Cahokia
Cree
k Tributary
near
Gree
n Creek
Trib
utar
y near
Onio
n Branch near St.
Cath
erin
e,Mo .
Traxler
Branch near Columbia,
Mo.
Brushy C
reek
near B
lairstown, Mo
.
Turkey Creek
near M
edin
a, Te
nn.
Wesl
ey Branch n
ear Walnut,
Miss.
Yandell
Branch near Kirbyville,
Mo.
Pike C
reek
Tri
bura
ry near
Poplar Bl
uff,
Mo
.
Jame
s Wolf Creek
Tributary
near
Mart
in L
ake
Tributary
at Si
don,
Moor
e's
Branch near
Wood
vill
e, Miss.
Hog
Fork Fishing
Creek
Tributary
Ten
Mile Cr
eek
Tributary
at
Ocmulgee Ri
ver
Tributary
near
Mulb
erry
Riv
er T
ributary
near
Winde r
, Ga .
Satilla
Rive
r Tr
ibut
ary
near
9 17
20 20
19
17 23
20 19
25 20
17
14
20 9 10
19 15
10 7 20
17
11
10
11
1 1
Rainfall-runoff
model
para
mete
rs
PSP
(in)
6.48
2.81
3.
31
1.92
1.41
1.08
1.28
3.49
4.49
3.
54
5.20
1.
83
1.96
.64
1.40
3.60
4.90
1.95
1.13
1.05
1.36
1.37
1.51
1.21
2.23
qft
KSAT
(in/h)
0.150
.095
.172
.051
.033
.050
.050
.055
.135
.082
.045
.043
.041
.032
.031
.056
.144
.061
.012
.013
.023
.047
.137
.103
.109
nAq
RGF
24.6
12.7
15.5
11.8
19.7
.S
.7
39.2
18.2
23.0
10.3
22.4
11.6
6.1
6.3
15.3
11.6
15.0 6.7
2.4
2.6
7.6
12.1
18.7
10.8
3.2
9n
/.
BMSM
(in)
3.14
1.46
8.
16
1.18
2.00
1.69
4.99
3.00
1.32
5.
75
6.64
5.
93
5.92
4.39
4.02
3.13
4.35
6.00
1.22
6.44
3.59
3.05
4.60
6.46
8.53
7 01
KSW
(h)
1.10 .80
.42
.57
1.75
1.50
7.50
.71
.32
.65
1.75
.5
0 .75
.94
.98
2.88
.66
.63
.82
2.70
.79
.70
4.18
4.44
1.89
i«
<;B
TC
(h)
1.33
1.83
1.
12 .98
1.25
4.17
3.42
.55
.27
1.10
1.33
.5
0 .50
1.50
2.25
4.13
.50
.33
.67
1.80
.74
.92
3.15
5.85
2.00
i n
sfi
Para
mete
rs u
sed
in map-model
estimating pr
oced
ure2
Area
(mi2
)
0.51
1.71
.22
.78
2.64
.83
6.30
.45
.44
4.90
1.04
.55
1.15
.22
7.87
2.
17
.33
.28
.29
.26
.21
.097
l.lt
3.23
2.68
f. IB
L (h)
1.77
1.72
.98
1.06
2.38
3.58
9.21
.98
.46
1.20
2.42
.75
1.00
1.67
2.
11
4.94
.91
.80
1.15
3.60
1.16
1.16
5.76
7.37
2.89
91 O
F
(in/h)
2.36 .46
1.08 .13
.12
.097
.27
.40
1.44
.43
.54
.14
.11
.051
.099
.32
1.24 .17
.020
.022
.054
.14
.52
.26
.27 i/.
°2 210
125
135
155
140
120
130
165
175
125
135
140
130
150
210
235
125
185
240
265
325
180
270
265
185
9B<;
°2S 620
560
600
510
540
500
510
650
725
750
660
500
650
630
640
650
640
750
675
750
1,100
650
850
860
670
onn
°100 875
800
875
725
750
650
700
1,000
1,150
1,20
0
1,000
700
975
925
900
925
925
1,05
0
950
1,100
1,60
0
925
1,25
0
1,25
0
950
1 97*
Table 3. Streamflow s
tati
ons
used
in study Continued
W
Map
code
, st
atio
n no
., name,
location,
and
numb
er of observed an
nual
floods
T06
T07
T08
T09
T10
Til
T12
T13
T14
T15
T16
T17
T18
T19
T20
T21
T22
T23
T24
T25
T26
T27
T28
T29
T30
T31
0231
7710
0231
7845
02342200
0237
1200
02
3872
00
0238
7560
02407900
0241
4800
0242
1300
02
4283
00
0243
7900
02
4502
00
0246
5205
02
4696
72
0247
2810
0248
3890
0248
9030
03
3136
00
0333
8100
0338
0300
0338
1600
0342
0360
0343
0400
03519650
0354
1100
0546
9750
Withlacoochee
Rive
r Tributary
near
Warr
ior
Cree
k Tr
ibut
ary
near
Sylvester, Ga
.
Phel
ps Cr
eek
near
Opelika, Al
a.
Beamer Creek
near
Sp
ring
Place, Ga
. Oo
thka
loog
a Creek
Trib
utar
y at
Paint
Creek
near
Marble
Valley,
Harb
uck
Creek
near Ha
ckne
yvil
le,
A1-
____ _
____ _
______ .
Tallatr.hee
Creek
near
Woods
Creek
near Ha
milt
on,
Ala.
Dors
ey Creek
near
Ar
kade
lphl
a, Ala.
Litt
le Okatuppa Cr
eek
near
Okat
oma
Creek
Tributary
No.
2
Yockanookany River
Tributary
near
McCo
ol,
Miss
. - -
Elmers Draw ne
ar Co
lumb
ia,
Miss
.
West
Fork Dr
akes
Cr
eek
Tributary
Sal
t Creek
Trib
utar
y ne
ar Catlin,
Til.
Dums Cr
eek
Trib
utar
y ne
ar lu
ka,
m_ _
___ _
________ _
Litt
le Wa
bash
Ri
ver
Trib
utar
y
Mud
Creek
Tributary
No.
2 ne
ar
Mill Creek
at Nolensville, Te
nn.
Litt
le Baker
Creek
near
Bitt
er Creek
near
Camp A
usti
n,
Tenn
. --
Elli
son
Creek
Tributary
near
Rner»7-M1o
Til
16 10
16
17 11 10
12 20
12 16
12
16
10 10 9 11
21 9 17
20
16 9 11 9 9 9n
Rain
fal
l- runoff
mode
l parameters
PSP
(in)
1.50
1.72
2.
28
5.72
.37
2.70
5.26
5.04
9.
11
1.54
2.
75
1.89
4.
97
2.16
1.02
1.68
3.24
3.30
1.71
3.33
3.49
2.30
1.26
5.98
2.98
9 fi<
;
KSAT
(in/h)
0.116
.038
.103
.154
.018
.103
.086
.057
.092
.059
.102
.046
.108
.031
.045
.022
.033
.088
.054
.040
.075
.076
.020
.139
.095
171
RGF
12.2
22.3
13.3
5.5
3.5
17.8
8.6
4.3
8.2
24.1
9.
3 10.7
13.9 9.2
18.0 6.1
16.7
15.3
24.9
16.3
16.1 9.2
13.3
13.9
7.7
Ifi
Q
BMSM
(in)
7.95
8.43
4.06
6.60
3.
26
7.16
6.50
2.83
5.
44
6.60
4.80
6.32
5.
24
5.51
3.60
5.52
1.82
5.50
1.92
1.50
1.17
3.04
3.
07
2.03
2.58
A nr»
KSW
(h)
5.30
5.60
4.75
6.50
2.07
2.38
4.50
2.50
4.
50
4.50
7.00
8.50
2.50
2.77
1.00
1.70
1.80 .67
2.54
.086
.214
2.75
.6
4
1.50
1.17 A7
TC
(h)
3.30
3.20
2.50
7.00
2.
23
2.00
4.23
2.75
5.
00
7.00
3.
50
3.50
1.
50
3.53
1.00
1.39
.65
.83
2.90
.75
.77
3.67
2.08
1.50
1.45
1 97
Para
mete
rs us
ed in
map-model
estimating procedure2
Area
(mi2)
0.86
1.64
7.
47
8.88
1.
29
3.56
13.5 6.70
10.5
14.6
14.1
13.0
3.56
4.35
.206
.34
.91
.95
2.20
.078
.156
2.28
12.0 3.65
5.53 9A
£ (h)
6.95
7.20
6.00
10.0
3.19
3.38
6.67
3.88
7.
00
8.00
8.75
10.3
3.25
4.54
1.50
2.40
2.
13
1.09
3.99
.46
.60
4.58
1.68
2.25
1.90
1 9K
F
(in/
h)
0.35 .18
.44
.90
.024
.59
.57
.28
.97
.26
.42
.15
.90
.11
.13
.055
.21
.55
.27
.26
.51
.27
.056
1.36
.38
79
«, 300
270
270
310
170
200
285
270
300
335
260
260
285
335
340
235
350
175
125
160
170
190
190
175
175
i An
C26 925
900
860
975
640
675
850
825
925
1,050
740
750
850
1,000
1,10
0
840
1,225
550
505
625
625
600
575
575
550
«;7«
;
°100
1,30
0
1,275
1,22
5 1,400
925
950
1,200
1,150
1,300
1,50
0 1,000
1,05
0 1,200
1,450
1,55
0
1,20
0 1,650
775
700
825
800
875
825
925
850
Ann
Tabl
e 3. Streamflow stations used in
study Continued
T32
T33
T34
T35
T36
T37
T38
T39
T40
g
T41
T42
T43
T44
T45
T46
T47
T48
T49
Map
code
,
numb
er
0549
5200
0555
8075
0557
7700
0559
5550
0682
1000
06896500
06908300
0692
5300
0702
8930
07
0355
00
07064500
0706
8200
0718
5500
07
2672
00
0728
3490
07287520
0729
0525
07294400
stat
ion
no., name,
loca
tion
, an
d of
ob
serv
ed an
nual
floods
Litt
le Creek
near
Breckenridge,
m _
. _ __
_ _
_
_Co
ffee
Creek
Tributary
near
Sang
amon
Ri
ver
Tributary
at
Mary
s Ri
ver
Trib
utar
y at
Ches
ter,
111.
Jenk
ins
Branch near Go
wer,
Mo
.
Thompson Br
anch
ne
ar Al
bany
, Mo.-
Trent
Bran
ch near Wa \rerly, Mo.
Prai
rie
Branch ne
ar
Turk
ey Creek
at Me
dina
, Tenn. -
Barnes Cr
eek
near
North
Pron
g Little Black
Rive
r
Crac
ker
Ditch
near
Po
ntot
oc,
Miss
. Ca
ney
Cree
k ne
ar Coffeeville,
Short
Creek
Tributary
near
White Oak
Creek
Trib
utar
y ne
arUt
ica,
Mi
ss.
Observers
Draw
near Do
loro
so,
W-fac
____. _ ._
_ .___ _ _ _ _______
Rain
fall
-run
off
model
parameters
20
20
20 15
25
17
20 20 9 19
25 17
24
19 21 9 11
20
PSP
(in)
1.41
3.88
2.70
4.17
1.
31
1.02
1.
97
4.03
2.20
2.16
2.36
12.40
1.71
1.
40
3.24
1.36
1.40
3.28
KSAT
(in/h)
0.07
0
.081
.095
.070
.060
.044
.059
.126
.048
.043
.036
.132
.048
.018
.035
.023
.028
.048
RGF
21.9
11.7
27.9
10.0
13
.8
5.8
14.9
14.2
16.3
14.5
8.
2
16.8
24.8
5.
9
19.3
16.7
15.1
11.2
BMSM
(in)
3.07
5.99
2.26
1.53
4.64
1.95
4.35
6.00
4.
88
4.28
2.02
6.70
3.52
4.
86
7.94
2.02
3.76
3.64
KSW
(h)
0.72
.106
1.44 .51
.75
2.00
.3
4
1.25
.7
5
.75
1.70
1.09
1.50
1.89 .61
1.05
1.33
.88
TC (h)
1.13
1.06
1.10 .49
.75
2.25
.75
.58
1.92
1.25
1.
25
1.00
1.75
1.
39
2.70
1.20
2.85
.28
Para
mete
rs us
ed in map-model
esti
mati
ng procedure2
Area
(m
i2)
1.45
.22
1.50 .65
2.72
5.58
.97
1.48
4.75
4.03
8.36
1.23
3.86
.23
1.97
1.49
1.36
.22
L (h)
1.29
.63
1.99 .76
1.13
3.
12
.71
1.54
1.
71
1.38
2.32
1.59
2.38
2.
58
1.96
1.65
2.76
1.02
F
(in/h)
0.27
.49
.75
.42
.18
.084
.24
.88
.22
.18
.12
2.89
.24
.040
.25
.076
.089
.25
°2 145
135
150
170
125
125
135
130
210
170
160
180
115
255
260
285
300
315
C25 550
520
540
725
750
750
650
550
630
725
690
750
590
710
750
860
960
1,025
C100
750
725
775
1,15
0 1,
150
1,20
0 1,
000
850
900
1,150
1,000
1,075
875
975
1,025
1,25
0
1,45
0
1,55
0
1Map co
des
D01
thro
ugh
D50
iden
tify
those
stat
ions
used to de
velo
p synthetic
annu
al floods.
Code
s T0
1 th
roug
h T49
iden
tify
th
e other
stat
ions
included in
th
e ev
alua
tion
of th
e map-model
esti
mati
ng pr
oced
ure.
'100
'-S
ee text for
a discussion of
th
e hydrograph shape
fact
or,
L, the
infiltration factor,
F, an
d th
e cl
imat
ic fa
ctor
s, C0
, <?oc
, an
d&
CiO
3Thi
s st
atio
n wa
s no
t in
clud
ed in
th
e ev
alua
tion
of th
e ma
p-mo
del
esti
mati
ng procedure
beca
use
the
reco
rd le
ngth
is
to
o sh
ort.
Table 4. Results of m
ulti
ple-
regr
essi
on an
alys
is fo
r ru
ral model
applications
2-yr
re
curr
ence
Rain
-gag
e lo
cati
on
Chicago, 111.
Peor
ia,
111.
Springfield, 11
1.
Atlantic Ci
ty,
N.J.
Balt
imor
e, Md
.
Kansas Ci
ty,
Mo.
St.
Louis, Mo
.
Louisville,
Ky.
Spri
ngfi
eld,
Mo.
Cair
o, 111.
Wyth
evil
le,
Va.
Nashville, Te
nn.
Knox
vill
e, Te
nn.
Char
lott
e, N.C.
Chat
tano
oga,
Tenn.
Gree
nvil
le,
S.C.
Litt
le Ro
ck,
Ark.
Columbia,
S.C.
Birm
ingh
am,
Ala.
Vick
sbur
g, Mi
ss.
Montgomery,
Ala.
Thomasville, Ga
.
Jacksonville,
Fla.
'kd.lOdV^UJ.Cl.y
J. J.Q.
New
Orle
ans,
La.
a
137
124
155
132
113
135
128
149
144
132
126 71.0
11
2 199 55.7
183
139
162
217
225
182
242
120
230
307
184
224
245
288
313
366
341
332
400
475
482
interval
Stan
dard
, ,
error
0.J
t>n
-0.7
2 -.
73
-.75
-.73
-.
63
-.73
-.74
-.
74
-.69
-.72
-.67
-.71
-.64
-.62
-.65
-.64
-.79
-.65
-.
70
-.65
-.
67
-.70
-.72
-.69
-.74
-.75
-.
71
-.64
-.70
-.70
-.75
-.74
-.
77
-.69
-.70
-0.56
-.56
-.
51
-.54
-.
55
-.55
-.58
-.53
-.52
-.
49
-.58
-.
67
-.58
-.
40
-.70
-.41
-.45
-.49
-.37
-.34
-.42
-.
37
-.58
-.
38
-.30
-.48
-.
40
-.41
-.30
-.32
-.29
-.35
-.35
-.31
-.27
-.28
^10
0.09
4 .090
.081
.086
.077
.091
.105
.087
.083
.074
.097
.102
.095
.061
.100
.060
.067
.080
.053
.049
.063
.063
.1
01
.061
.045
.078
.062
.074
.050
.054
.050
.062
.062
.055
.0
46
.051
Perc
ent
21.9
21.0
18
.7
20.0
17
.7
21.1
24.4
20
.3
19.2
17.2
22.6
23.7
22.0
14
.0
23.1
13.8
15
.4
18.5
12.2
11.4
14.6
14
.5
23.5
14
.1
10.4
18.1
14
.3
17.0
11.4
12
.4
11.4
14
.3
14.4
12.6
10.5
11.7
25-y
r re
curr
ence
interval
a 738
504
543
574
783
565
692
527
698
469
615
437
565
734
413
615
598
531
652
685
685
738
542
753
830
665
815
685
900
998
885
1240
95
2 1140
1500
1340
Stan
dard
L
-L. er
ror
fcj
00
-0.73
-.70
-.70
-.75
-.63
-.69
-.69
-.74
-.69
-.72
-.71
-.75
-.68
-.65
-.
81
-.73
-.74
-.76
-.66
-.75
-.70
-.65
-.73
-.76
-.67
-.74
-.72
-.63
-.59
-.65
-.72
-.68
-.68
-.68
-.64
-0.2
3 -.
28
-.22
-.25
-.21
-.30
-.24
-.26
-.21
-.25
-.30
-.36
-.24
-.19
-.33
-.21
-.
23
-.24
-.19
-.18
-.25
-.
17
-.29
-.20
-.16
-.24
-.20
-.18
-.16
-.14
-.19
-.11
-.15
-.11
-.11
-.10
*VM
0.05
7 .053
.040
.043
.050
.066
.053
.048
.037
.029
.065
.071
.048
.038
.065
.034
.0
40
.041
.033
.025
.043
.032
.052
.035
.037
.040
.032
.030
.032
.031
.033
.034
.025
.030
.031
.030
Perc
ent
13.1
12
.1
9.2
9.9
11.5
15.3
12.3
11.0
8.6
6.7
15.0
16.3
11.1
8.
6 14
.9 7.9
9.1
9.5
7.6
5.6
10.0
7.3
12.0
8.
1 8.
5
9.3
7.5
7.0
7.3
7.1
7.5
7.7
5.6
6.9
7.2
6.9
100-yr recurrence in
terv
al
a
Stan
dard
h
h er
ror
log~
n Pe
rcen
t
1230
67
8 77
6 85
7 1600 783
1110
740
1110
679
957
759
889
1090
76
0
837
966
729
904
940
973
953
755
1010
1090 935
1190
85
9 1190
1250
1100
17
10
1230
14
30
2130
18
00
-0.75
-.67
-.68
-.75
-.
64
-.66
-.66
-.
75
-.65
-.71
-.67
-.
76
-.66
-.
66
-.88
-.77
-.
75
-.73
-.66
-.
77
-.71
-.
65
-.72
-.76
-.
66
-.74
-.71
-.60
-.
56
-.61
-.72
-.
66
-.65
-.63
-.61
-.54
-0.14
-.21
-.
12
-.17
-.
10
-.24
-.14
-.18
-.12
-.17
-.22
-.
26
-.14
-.
14
-.21
-.18
-.17
-.17
-.16
-.
16
-.22
-.14
-.21
-.18
-.14
-.18
-.
16
-.12
-.14
-.12
-.19
-.
06
-.10
-.
07
-.08
-.06
0.047
.043
.035
.037
.049
.063
.0
43
.049
.031
.034
.054
.062
.042
.039
.057
.038
.041
.032
.037
.028
.051
.042
.043
.045
.045
.039
.0
38
.036
.035
.037
.043
.038
.030
.040
.036
.032
10.8
9.
9 8.
0 8.
4 11
.4
14.5
9.3
11.3
7.1
7.8
12.5
14
.3
9.6
8.9
13.1 8.8
9.5
7.3
8.5
6.5
11.7
9.
7 9.
9 10.3
10.3 8.9
8.7
8.3
8.0
8.5
9.8
8.6
6.8
9.2
8.3
7.3
Tabl
e 5. Summary and
comparison of
mu
ltip
le-r
egre
ssio
n (m-r)
and
direct-search
(d-s
) determin
ations for
the
coef
fici
ent,
a.
2-yr
recurrence interval
Rain
-gag
e lo
cati
on
Dub uq ue
, I o
wa
Chic
ago,
111.
Peor
ia,
111.
-
Springfield, 111.
Atla
ntic
City,
N.J.
Baltimore, Md
.
Kans
as Ci
ty,
Mo.
Colu
mbia
, Mo
.
St.
Louis, Mo
.
Loui
svil
le,
Ky.
Rich
mond
, Va.
Ly nch
bu rg
, Va .
Spri
ngfi
eld,
Mo.
Cairo, 11
1.
Wyth
evil
le,
Va.
Nash
vill
e, Tenn.
Knox
vill
e, Tenn.
Char
lott
e, N.
C.
Memphis, Tenn.
Chattanooga, Tenn.-
Gree
nvil
le,
S.C.
Litt
le Ro
ck,
Ark.
Columbia,
S.C.
Atla
nta,
Ga.
Birmingham,
Ala.
Augu
sta,
Ga
.
Maco
n, Ga.
Shreveport,
La.
Vicksburg, Mi
ss.
Montgomery,
Ala.
Meri
dian
, Mi
ss.
S a van
nah
, Ga .
Thomasville, Ga
.
Jack
sonv
ille
, Fl
a.-
Pensacola, Fla.
New
Orle
ans,
La.
am-
v
137
124
155
132
113
135
128
149
144
132
126 71.0
112
199 55.7
183
139
162
217
225
182
242
120
230
307
184
224
245
288
313
366
341
332
400
475
482
d-s
140
121
151
129
122
138
131
149
140
129
129 77.2
115
199 63.1
178
139
147
212
204
182
236
123
213
285
179
203
239
281
298
348
324
316
362
463
470
Stan
dard
log
m-r
0.094
.090
.081
.086
.077
.091
.105
.087
.083
.074
.097
.102
.095
.061
.099
.060
.067
.080
.053
.049
.063
.063
.103
.061
.045
.078
.062
.073
.049
.054
.049
.062
.062
.055
.046
.051
10
d-s
0.09
6.0
92.0
81.0
86.080
.091
.108
.089
.083
.076
.100
.105
.095
.067
.106
.068
.076
.085
.065
.073
.068
.066
.102
.068
.061
.079
.069
.073
.066
.061
.057
.065
.064
.063 048
.051
error
Perc
ent
m-r
21.9
21.0
18.7
20.0
17.7
21.1
24.4
20.3
19.2
17.2
22.6
23.7
22.0
14.0
23.1
13.8
15.4
18.5
12.2
11.4
14.6
14.5
24.0
14.1
10.4
18.1
14.3
17.0
11.4
12.4
11.4
14.3
14.4
12.6
10.5
11.7
d-s
22.2
21.0
18.8
19.8
18.3
21.1
25.0
20.3
19.1
17.5
23.2
24.1
22.0
15.4
24.4
15.5
17.4
19.6
14.9
16.6
15.5
15.1
23.7
15.6
13.9
18.1
15.7
16.7
15.1
13.9
13.1
14.9
14.7
14.3
10.9
11.7
am-r 738
504
543
574
783
565
692
527
698
469
615
437
565
734
413
615
598
631
652
685
685
738
542
753
830
665
815
685
900
998
885
1240 652
1140
1500
1340
25-y
r
d-s
720
504
504
533
844
579
710
489
698
424
631
448
565
753
364
571
555
430
652
604
702
738
529
699
809
649
795
685
995
1020 885
1240 952
1110
1660
1520
recu
rren
ce interval
Stan
dard
log
m-v
0.05
7.052
.040
.043
.050
.066
.053
.048
.037
.029
.065
.071
.048
.037
.065
.034
.039
.041
.033
.024
.043
.032
.052
.035
.037
.040
.032
.030
.032
.031
.033
.034
.024
.030
.031
.030
10
d-s
0.058
.052
.050
.049
.056
.069
.054
.053
.038
.045
.071
.076
.049
.042
.077
.044
.045
.054
.041
.047
.044
.040
.053
.042
.042
.043
.034
.044
.049
.037
.033
.034
.024
.034
.043
.050
erro
r
Perc
ent
m-r
13.1
12.1 9.2
9.9
11.5
15.3
12.3
11.0 8.6
6.7
15.0
16.3
11.1 8.6
14.9 7.9
9.1
9.5
7.6
5.6
10.0 7.3
12.0 8.1
8.5
9.3
7.5
7.0
7.3
7.1
7.5
7.7
5.6
6.9
7.2
6.9
d-s
13.2
11.9
11.5
11.2
12.7
16.0
12.4
12.1 8.6
10.4
16.7
17.6
11.2 9.5
17.6
10.2
10.3
12.2 9.5
10.8
10.3 9.1
12.2 9.6
9.6
9.8
7.7
10.1
11.3 8.4
7.5
7.7
6.3"
7.7
9.8
11.5
am-r
1230 678
776
857
1600 783
1110 740
1110 679
957
759
889
1090 760
837
966
729
904
940
973
953
755
1010
1090 935
1190 859
1190
1250
1100
1710
1230
1430
2130
1800
100-
yr
d-s
1170 678
720
775
1770 865
1140 670
1170 614
1060 759
889
1150 607
757
896
660
904
851
998
953
736
961
1120 912
1220 903
1420
1380
1130
1800
1260
1500
2470
2140
recu
rren
ce interval
Stan
dard
log
m-r
0.047
.043
.035
.037
.049
.063
.040
.049
.031
.034
.054
.062
.042
.039
.057
.038
.041
.032
.037
.028
.051
.042
.043
.044
.044
.039
.038
.036
.035
.037
.043
.037
.030
.040
.036
.032
10
d-s
0.050
.044
.059
.046
.057
.069
.046
.060
.036
.053
.065
.070
.049
.041
.092
.050
.048
.046
.039
.043
.059
.046
.043
.049
.046
.041
.040
.057
.067
.050
.047
.040
.036
.047
.064
.071
error
Perc
ent
m-v
10.8 9.9
8.0
8.4
11.4
14.5 9.3
11.3 7.1
7.3
12.5
14.3 9.6
8.9
13.1 8.8
9.5
7.3
8.5
6.5
11.7 9.7
9.9
10.3
10.3 8.9
8.7
8.3
8.0
8.5
9.8
8.6
6.8
9.2
8.3
7.3
d-s
11.4
10.1
13.5
10.5
13.0
15.7
10.6
13.8 8.2
12.2
14.9
15.9
11.1 9.5
21.2
11.5
11.0
10.6 9.0
9.8
14.0
10.6 9.9
11.3
10.5 9.4
9.3
13.0
15.4
11.4
11.1 9.1
7.9
10.8
14.7
16.2
Table 6. Results of
the
dire
ct-s
earc
h determinations fo
r th
e co
effi
cien
t, d.
t for
effe
ct of
Im
perv
ious
ar
ea
Rain-gage
location
Peor
ia,
111.
Spri
ngfi
eld,
11
1.
Atlantic Ci
ty,
N.J.
Balt
imor
e, Md
.
Kansas Ci
ty,
Mo.-
Spri
ngfi
eld,
Mo
.
Nash
vill
e, Tenn.
Knox
vill
e, Te
nn.
Char
lott
e, N.C.
Chat
tano
oga,
Tenn.
Gree
nvil
le,
S.C.
Litt
le Ro
ck,
Ark.
Columbia,
S.C.
Birmingham,
Ala.
Shreveport,
La.
Vick
sbur
g, Mi
ss.
Montgomery,
Ala.
Thom
asvi
lle,
Ga
.
Jack
sonv
ille
, Fl
a.
New
Orleans, La.
2-yr
recurrence in
terv
al
a 140
121
151
129
122
138
131
149
140
129
129 77.2
11
5 19
9 63.1
178
139
147
212
204
182
236
123
213
285
179
203
239
281
298
348
324
316
362
463
470
d 5.61
5.70
4.
78
5.12
5.
61
5.29
5.72
4.89
4.98
4.63
5.81
7.80
6.02
3.57
8.66
3.82
4.18
4.
78
3.30
3.20
3.72
3.26
5.85
3.36
2.72
4.26
3.
68
3.43
2.
73
2.83
2.57
3.
02
3.01
2.69
2.35
2.
42
*"»
0.052
.056
.050
.0
52
.050
.052
.0
58
.053
.050
.050
.054
.055
.052
.046
.060
.047
.051
.057
.0
48
.051
.047
.044
.0
55
.046
.045
.050
.047
.046
.051
.042
.041
.044
.0
44
.047
.036
.037
Perc
ent
12.1
12.8
11.6
12.0
11.6
11.9
13
.4
12.2
11.5
11.6
12.4
12.8
12.0
10.7
13.9
10.9
11.7
13.1
11.1
11
.9
10.9
10.2
12.6
10
.6
10.3
11.5
10.8
10.5
11
.8
9.8
9.4
10.2
10.2
10.9
8.4
8.6
25-y
r recurrence In
terv
al
a 720
504
504
533
844
579
710
489
698
424
631
448
565
753
364
571
555
480
652
604
702
738
529
699
809
649
795
685
995
1,02
0
885
1,240
952
1,110
1,660
1,520
d 1.91
2.26
2.15
2.19
1.73
2.60
2.12
2.39
1.91
2.29
2.42
2.
68
2.01
1.
84
2.88
1.92
2.
15
2.21
1.83
1.89
2.16
1.81
2.45
1.99
1.72
2.11
1.80
2.06
1.60
1.
56
1.75
1.
46
1.60
1.51
1.36
1.35
^10
0.047
.041
.038
.038
.048
.046
.042
.037
.033
.032
.052
.052
.041
.040
.061
.035
.038
.038
.041
.033
.037
.040
.0
40
.037
.033
.034
.032
.041
.047
.039
.032
.035
.028
.029
.0
43
.052
Perc
ent
10.7
9.6
8.7
8.7
11.0
10.7
9.
7 8.
4 7.
6 7.5
12.1
11
.9
9.4
9.1
14.1 8.1
8.8
8.8
9.4
7.7
8.6
9.2
9.3
8.5
7.7
7.9
7.5
9.5
10.9
8.
9
7.5
8.0
6.4
6.7
10.0
12.0
100-
yr re
curr
ence
Interval
a
1,170
678
720
775
1,77
0
865
1,140
670
1,17
0 614
1,06
0 75
9 88
9 1,150
607
757
896
660
904
851
998
953
736
961
1,120
912
1,22
0 903
1,420
1,380
1,130
1,800
1,260
1,500
2,470
2,14
0
d 1.43
2.09
1.65
1.94
1.11
2.25
1.76
2.08
1.
57
1.76
1.82
2.
00
1.53
1.50
2.06
1.74
1.
70
1.88
1.75
1.72
2.19
1.
82
2.16
1.89
1.60
1.86
1.64
1.
85
1.59
1.55
1.80
1.29
1.
46
1.32
1.26
1.25
*»»
0.050
.047
.041
.034
.0
57
.049
.0
43
.038
.035
.032
.058
.0
55
.048
.045
.078
.039
.043
.031
.045
.033
.045
.048
.038
.040
.036
.031
.041
.0
53
.058
.052
.041
.041
.035
.046
.057
.070
Perc
ent
11.5
10.8
9.5
7.8
13.3
11.3
10.0
8.8
8.0
7.3
13.5
12.8
11.1
10.5
18.0 9.1
9.8
7.2
10.3
7.7
10.4
11.1
8.8
9.2
8.3
7.2
9.4
12.3
13.3
12.0 9.4
9.4
8.0
10.6
13.3
16.2
Tab
le 7
. S
tan
dard
err
ors
fo
r th
e si
ngle
-coeff
icie
nt,
sy
nth
eti
c
flood re
lati
on
Rain
-gag
e lo
cati
on
Pam-l a
T-11
_______
_______ ________
Spri
ngfi
eld,
111.
Atla
ntic
Ci
ty,
N.J.
Balt
imor
e, Md.
Spri
ngfi
eld,
Mo.
Cairo, 11
1.
Wyth
evii
le,
Va.
Krioxv.llle, Te
nn.
Char
lott
e, N.C.
Greenville,
S.C.
Colu
mbia
, S.C.
Stan
dard
er
ror
in pe
rcen
t fo
r va
riou
s
2-yr
0
22.2
21.0
1O Q
i Q
aTO
T
25.0
_ ._ __
on
"5
19.1
17.5
23.2
24.1
____
07 n
15.4
j. _?
j
17.4
19.6
14.9
16.6
15.5
15.1
_____
oo 7
15.6
13.9
1 Q
1
15.7
J.U
/
15.1
13.9
13.1
14.9
14.7
Uo
10.9
11.7
recu
rren
ce in
terv
al
15 15.2
14.6
13.2
13.7
13.1
14.1
16.3
13.9
13.2
13
.6
15.5
14.9
14.0
12
.4
15.8
12.7
14.4
14.6
12.9
14.7
12.8
11
.6
15.0
12
.4
11.9
13.3
12.4
22.0
13
.4
11.2
10.8
11.6
11.7
12.0
9.
3 9.
6
30 13.3
12.4
10
.9
11.0
10.9
11.2
13.5
11.4
10.5
12.1
13.3
11
.7
11.4
10.4
12
.7
11.1
13.3
12.4
11.6
13.7
11.0
9.7
12.0
10
.5
10.4
10.9
10
.2
9.7
12.2
9.
3
9.0
10.4
10.3
10.6
8.
1 8.
3
45 12.0
11
.0
9.5
9.1
9.4
9.6
12.0
9.
8 8.
7 11
.1
12.1
9.
6 9.5
8.8
10.8 9.7
12.7
10.8
10.6
13.1 9.4
8.1
10.3
9.0
9.1
9.5
8.4
8.5
11.1
7.
6
7.5
10.2
9.6
9.8
7.1
7.4
25-yr
0 13.2
11.9
11.5
11
.2
12.7
16.0
12.4
12.1
8.6
10.4
16.7
17.6
11.2
9.5
17.6
10.2
10.3
12.2
9.5
10.8
10.3
9.1
12.2
9.
6 9.
6
9.8
7.7
10.1
11
.3
8.4
7.5
7.7
6.3
7.7
9.8
11.5
leve
ls of
im
perv
ious
ness
recu
rren
ce interval
15 12.2
11.0
10.0
9.
8 12
.5
14.0
10.8
9.5
8.5
8.6
14.7
13
.6
11.1
9.7
15.3 9.6
9.6
10.2
10
.0
8.9
9.9
9.4
10.7
8.
9 8.
5
8.6
8.1
10.0
11.3
9.3
7.8
8.0
6.8
7.0
10.1
12
.0
30 10.2
9.
0 8.
8 8.
3 10
.7
13.3
10.0
8.1
7.5
7.9
13.4
11.8
9.6
9.4
14.2 9.2
8.7
8.6
10.5
8.
4
9.2
9.7
10.0
8.
4 8.
0
7.8
7.7
9.7
11.1
9.
7
7.7
8.5
7.5
7.2
10.2
12.3
45 8.7
7.7
7.5
7.1
9.3
13.1
9.7
7.5
6.6
6.9
13.5
12
.0
8.0
8.8
13.8 8.2
7.8
7.3
10.3
7.
9
9.1
9.6
9.9
8.0
7.5
7.4
7.0
9.4
11.0
9.
5
7.3
8.8
7.6
7.7
10.1
12
.4
100-yr
0 11.4
10.1
13.5
10.5
13.0
15.7
10.6
13.8
8.
2 12
.2
14.9
15.9
11.1
9.5
21.2
11.5
11
.0
10.6
9.0
9.8
14.0
10.6
9.
9 11
.3
10.5 9.4
9.3
13.0
15.4
11
.4
11.1
9.
1 7.9
10.8
14
.7
16.2
recu
rren
ce in
terv
al
15 13.1
11.6
11
.3
8.6
15.4
15.0
10.7
9.
6 8.
3 8.
5
15.3
14
.5
13.0
10.3
19.1
10.0
10
.6
8.1
9.7
8.2
14.5
10
.5
9.9
10.0
9.0
8.0
9.6
12.6
14
.2
11.8
10.3
8.
8 7.2
10.2
14.0
16
.4
30
12.3
11.0
11.3
7.6
14.6
14.8
10
.8
8.3
8.2
10.2
13.8
12
.4
13.1
11.2
17
.7 9.6
10.1
7.8
10.9
7.9
15.0
11.7
10.1
9.8
7.9
7.6
10.0
12.5
13.8
12.6
10.7
9.
7 8.
7 10
.9
13.9
16
.5
45 10.9
10
.5
10.9
6.8
13.4
14.4
10
.9
8.0
7.8
10.6
12.7
11
.3
12.1
11.0
16.7 8.8
9.0
7.2
11.1
7.
7
15.3
12
.5
9.9
9.8
7.0
7.4
9.9
12.5
13.4
12.4
10.9
10
.5
9.4
12.3
13.8
16.3